Sicc Media // Break the Mold \\ DARE TO BE EXTRAORDINARY.

Blog

  • AI-Driven Proposal and Bid Generators

    AI-Driven Proposal and Bid Generators

    AI is good at the words around the number. It should never be allowed near the number.

    TL;DR AI drafts the scope narrative from your site notes. Pricing comes from your own calculator, always. Inject the case study that matches their situation. Assemble the contract from approved clauses, not generated text.

    The hard boundary

    AI drafts Never AI
    Scope of work narrative The price
    Explanation of the approach Contract terms
    Why option A over option B Compliance statements
    Case study framing Warranty wording
    Covering note Any statistic or claim
    Summary of findings Timescales you have not confirmed

    Everything in the right column is a legal or commercial commitment. Generated text there is a liability, not a time saving.

    Scope narrative from site notes

    The workflow

    1. Structured intake on site. Job type, findings, measurements, condition flags, access notes, customer priority.
    2. AI drafts the narrative using your template and your previous approved proposals as the style reference.
    3. You insert the price from your own calculator.
    4. You read every line and correct it.
    5. Send.

    The time saving is real: twenty to forty minutes per proposal on the descriptive sections, which are the parts you currently rewrite from scratch and resent writing.

    Feed it your own past proposals. The output quality difference between a generic prompt and a prompt containing three of your own approved proposals is enormous.

    Prompt structure that works

    Give it constraints, not creative licence.

    Draft the scope section of a proposal using the notes below and matching the tone of the three attached examples.

    Rules:
    – No prices. Leave [PRICE] placeholders.
    – No timescales other than those in the notes.
    – No statistics, percentages or research claims of any kind.
    – No superlatives about our company.
    – British spelling.
    – Under 400 words.
    – If information is missing, write [NEED: ...] rather than assuming.

    Notes: [paste intake]

    The [NEED: ...] instruction is the most useful line in that prompt. It converts the model’s tendency to fill gaps with plausible invention into a visible flag you can act on.

    Dynamic pricing, from your own logic

    Build the calculator separately and deterministically.

    • Rate tables per job type.
    • Quantity and measurement inputs.
    • Condition multipliers.
    • Materials at current cost plus your margin.
    • Travel and access adjustments.
    • Rounding rules.

    Then merge the calculated figure into the document. The AI never sees the pricing logic and never generates a number.

    Review the rate tables quarterly. Material costs move and a calculator with last year’s numbers produces confident, precise, wrong quotes.

    Case study injection

    Match the case study to their situation, automatically.

    Build a small library, tagged by job type, property type, and problem.

    Selection rule: closest match on job type first, then property type.

    Similar job: A 1930s semi on the other side of town, same boxed-in pipework problem. Two days, no disruption to the rest of the house. Photos below, and the owner is happy to be contacted if you want a reference.

    Offering the reference contact is the strongest element and it costs nothing if you have permission on file.

    Never let AI write the case study. Fabricated project examples are a serious problem. These come from your real completed jobs, with permission, and get injected as fixed blocks.

    Contract assembly

    Modular, from approved clauses. Never generated.

    How it works

    • A library of clauses, each reviewed by a solicitor once.
    • Rules selecting which apply. Job type, value band, customer type, jurisdiction.
    • Assembly into the document.
    • No generation, no paraphrasing, no summarising of legal text.

    Version and date every clause. Know which version each customer signed.

    Review the library annually with the same solicitor. Cheaper than reviewing every contract and far safer than generating them.

    The proposal document

    Structure

    1. The outcome, in one line. What they get.
    2. What we found, from the site visit. Photographs.
    3. What we propose, the AI-drafted scope, corrected.
    4. Options, if relevant. Two or three, never five.
    5. Price, broken into components.
    6. What is not included, explicitly.
    7. Timeline and availability.
    8. Similar job, with photos.
    9. Accept button.
    10. Terms, assembled.

    Accept button at the top and the bottom. Not only after the terms.

    Guardrails to build in

    • A verification checklist before send. Price checked, no invented claims, dates correct, name correct.
    • Search for [NEED: and [PRICE] before sending. Never send a document containing a placeholder.
    • A second reader for anything above a value threshold.
    • A log of what was sent, so a dispute can be resolved from record.

    The most common failure is sending a draft. Build the placeholder check into the send step mechanically.

    Measure it

    • Time from site visit to proposal sent.
    • Proposal acceptance rate, before and after. If it drops, the generated text is reading as generic.
    • Time spent per proposal, including editing.
    • Errors caught at review. Rising means the prompt needs tightening.
    • Questions from customers about proposal content, which indicate unclear sections.

    Build the structured site intake form first. AI-assisted drafting is only as good as the notes it works from, and better notes improve your proposals whether or not you ever automate the writing.

    Need a pro to build it? [BOOK A CALL]

  • Automating Subcontractor Management and Onboarding

    Automating Subcontractor Management and Onboarding

    Every subcontractor relationship carries a compliance obligation, and the one you forget about is the one that turns up in an insurance claim.

    TL;DR Nobody works until the document pack is complete. Track every expiry date with automatic reminders at 60, 30 and 7 days. Pay fast and consistently. Score performance on data you already collect.

    The document pack

    Define it once and never accept a partial set.

    Typically required

    • Tax documentation appropriate to your jurisdiction. In the US this means a W-9 for a US-based contractor.
    • Public liability insurance certificate, with the expiry date.
    • Employer’s liability, if they have their own staff.
    • Professional indemnity, where the work warrants it.
    • Trade licences and certifications, with expiry dates.
    • Vehicle insurance covering business use.
    • Signed subcontractor agreement.
    • Bank details for payment, verified.
    • Health and safety documentation, where required.

    Requirements vary by jurisdiction and by trade. Confirm your list with an accountant and an insurance broker rather than copying a generic one, and revisit it annually.

    No documents, no work

    This is the rule and it has to be absolute.

    How it fails in practice: a job comes in, a subcontractor is available, the paperwork “will follow”, and it never does. Then something goes wrong.

    Automate the gate

    1. Subcontractor added to the system, marked “pending”.
    2. Automatic email with the document checklist and upload links.
    3. Reminders at 3, 7 and 14 days.
    4. Each document reviewed and approved individually.
    5. Status only flips to “active” when all are approved.
    6. Dispatch cannot assign a pending subcontractor. Enforced by the system, not by memory.

    The system enforcement is the important part. A policy that relies on somebody remembering during a busy week is not a control.

    Insurance expiry tracking

    The highest-risk item and the easiest to automate.

    On approval, record the expiry date as a structured date field.

    Then automate

    Days before expiry Action
    60 Reminder to the subcontractor
    30 Second reminder, copy internal
    7 Escalation, internal alert
    0 Status flips to inactive automatically

    The automatic status change at expiry is essential. Without it, reminders get ignored and expired cover is discovered after the incident.

    Working with an uninsured subcontractor can invalidate your own cover. This is not administrative tidiness, it is the risk that ends businesses.

    Payout triggers

    Fast, consistent payment is your competitive advantage in retaining good subcontractors.

    Automate the chain

    1. Job marked complete by the subcontractor, with photos and signature.
    2. Internal quality check, if required.
    3. Payment record generated automatically from the agreed rate.
    4. Approval, one tap.
    5. Payment scheduled in the next run.
    6. Remittance sent automatically.

    Set a fixed payment day and hold it. Predictability matters more to subcontractors than speed. The best ones prioritise the people who pay when they said they would.

    Self-billing, where permitted, removes the invoice chase entirely. You generate the payment record from the completed job. Check the tax rules that apply before adopting it.

    Rates and scope

    Document per subcontractor

    • Rate structure. Day rate, hourly, per job, percentage.
    • What is included. Materials, travel, waste disposal.
    • Payment terms.
    • Cancellation and standby terms.
    • Warranty obligations on their work.

    The warranty clause is the one most often missing. If a subcontractor’s work fails in eight months, who returns and who pays? Agree it in writing at onboarding, not at the point of failure.

    Performance scorecards

    Score on data you already collect. Do not build a survey.

    Metric Source
    On-time arrival Time tracking
    Job completion rate Job records
    Callback rate Complaints and return visits
    Photo and record completeness Job records
    Customer feedback Post-job survey
    Documentation currency Compliance status

    Review quarterly. Share the scorecard with them, because most performance problems are informational rather than attitudinal.

    Act on it

    • Top performers get first refusal on work. Say so explicitly, because it drives behaviour.
    • Mid performers get the specific feedback.
    • Consistent poor performers get one clear conversation, then removal.

    Volume of work is your strongest lever. Better than rates, and it costs you nothing.

    Classification risk

    Worth stating plainly. How you treat a subcontractor affects whether authorities regard them as a contractor or an employee, and the tests vary by jurisdiction and are applied strictly in many places.

    Factors commonly considered

    • Control over how and when the work is done.
    • Whether they can send a substitute.
    • Whether they work for others.
    • Who provides tools and materials.
    • How integrated they are into your business.

    Get advice specific to your jurisdiction. Misclassification is expensive and retrospective. This article is not that advice.

    Onboarding beyond compliance

    The paperwork is the gate. The onboarding is what makes them work well.

    • Your standards, in writing. What a completed job looks like.
    • Access to your training videos, at least the customer-facing ones.
    • Introduction to the systems they will use.
    • Named contact for questions.
    • A first job with a check-in, not a cold start.

    Subcontractors represent you to the customer. The customer does not know or care about the contractual distinction.

    Measure it

    • Time from first contact to fully approved. Should be days, not weeks.
    • Percentage with current documentation. Must be 100.
    • Expired documents caught before expiry, versus after.
    • Payment on agreed date, percentage.
    • Callback rate, subcontracted versus in-house work.

    Build the insurance expiry reminders this week. It is a date field and three scheduled messages, and it addresses the single largest uninsured risk most growing service businesses carry.

    Need a pro to automate it? [BOOK A CALL]

  • Streamlining Field Technician Mobile Workflows

    Streamlining Field Technician Mobile Workflows

    Paper job sheets get lost, arrive three days late and are illegible. The information you need to run the business is sitting in a van door pocket.

    TL;DR Count the taps before you buy. Offline mode must genuinely work. Invoice from site, which transforms cash flow. Roll out with one technician before all of them.

    Count the taps

    The single best evaluation test, and almost nobody does it.

    Complete a full job in the app, on a phone, and count every tap and every screen.

    • Under 15 taps for a routine job: good.
    • 15 to 25: workable, expect some resistance.
    • Over 25: it will not be used properly and your data will be incomplete.

    Do this on a phone, not a tablet, not a laptop. Most technicians use a phone regardless of what the business bought.

    Do it wearing work gloves if that is realistic for the trade. Small touch targets fail in real conditions and no demo will show you this.

    Offline is non-negotiable

    Plant rooms, basements, lofts, rural properties, lift shafts. Signal fails routinely.

    Test properly

    1. Put the phone in airplane mode.
    2. Complete an entire job, including photos.
    3. Turn connectivity back on.
    4. Verify everything synced, including full-resolution photos.

    Failure modes to watch for

    • Photos queue but never upload.
    • The app appears to work but discards data on reconnect.
    • Sync conflicts if the office edited the same job.
    • A spinner that never resolves, forcing a force-quit that loses the entry.

    If offline mode fails this test, the platform is unsuitable, whatever else it does well.

    The mobile job sheet

    Structure it in the order the work actually happens.

    1. Job details and customer history. What was done last time, visible immediately.
    2. Arrival. Tap to start, which triggers the on-the-way text and the time log.
    3. Before photos.
    4. Findings. Fixed list plus free text.
    5. Work done. Fixed list.
    6. Parts used. Selected or scanned.
    7. After photos.
    8. Recommendations, for the next visit or a follow-up quote.
    9. Customer signature.
    10. Complete, which triggers invoicing and the feedback sequence.

    Fixed lists wherever possible. Free text does not aggregate, cannot be reported on, and takes longer to enter on a phone.

    Required fields on the things that matter, and nowhere else. Over-requiring produces junk entries as people work around it.

    Photos

    The most valuable data captured on site.

    • Minimum counts enforced. Two before, two after, at least.
    • Automatic tagging to the job, with timestamp and location.
    • Sensible compression. Full resolution on mobile data is slow and expensive.
    • Annotation on device, to circle the problem.
    • Same angle before and after.

    Photos are your defence in a dispute, your marketing content, and your customer communication. Treat the requirement as non-negotiable.

    Digital signatures

    On the phone, on site, before leaving.

    What it needs to capture

    • Signature and printed name.
    • Timestamp and location.
    • What exactly is being signed for. Work completed, or authorisation for extra work.
    • Automatic copy to the customer.

    The extra-work authorisation is the one that saves money. A signed on-screen approval for additional work found on site prevents the most common invoicing dispute in the trades.

    Invoicing from site

    The highest-impact change in this whole article.

    Traditional cycle: job done, paperwork returns to office, invoice raised days later, payment terms start, payment arrives weeks after the work.

    On-site cycle: job done, invoice generated on the phone, emailed or texted immediately, payment link included, frequently paid before the van leaves the street.

    What it changes

    • Days sales outstanding falls substantially.
    • Disputes drop, because the customer is standing there while the invoice is explained.
    • Admin time collapses. No re-keying.
    • Payment on the spot becomes normal rather than awkward.

    Enable card payment on the device. Tap to pay on a phone removes the last friction point.

    Rolling it out

    One technician first. The one most likely to engage, not the one who most needs fixing.

    1. Pilot with one person for two weeks. Real jobs.
    2. Fix what they hit. There will be things.
    3. Have them train the next two. Peer training beats management training substantially.
    4. Roll out with paper running in parallel for a fortnight.
    5. Set a hard paper cut-off date and hold it.

    The parallel period must end. Indefinite dual running means nobody commits to the app.

    Expect a temporary productivity dip. Say so in advance so it does not read as failure.

    Handling resistance

    Most resistance is reasonable and specific.

    • “It takes longer.” Often true at first. Count the taps and fix the genuinely slow parts.
    • “My phone battery dies.” Real. Provide chargers, and check the app’s background behaviour.
    • “I cannot see it in sunlight.” Real. Nothing you can do about the screen, but you can reduce how much reading is required.
    • “It is surveillance.” Address directly and honestly. Say what is tracked and what is not.

    Listen to the specifics. Blanket resistance is rare. Specific complaints are usually accurate and fixable.

    Measure it

    • Time from job completion to invoice sent. Should be minutes.
    • Days sales outstanding, before and after.
    • Jobs with complete records. Photos, parts, signature. Target near 100 percent.
    • Admin hours per week in the office.
    • Taps per job, revisited after a few months as the workflow settles.

    Time yourself completing one full job in your current app, on a phone, counting taps. If it is over twenty-five, fix that before rolling anything out further.

    Need a pro to set it up? [BOOK A CALL]

  • Centralizing Customer Communications into One Inbox

    Centralizing Customer Communications into One Inbox

    Messages get missed because they arrive in six places and no single person is responsible for any of them.

    TL;DR One inbox, every channel, threaded by customer. Assign every conversation to a named person. Track first-response time as your main metric. Tag automatically, respond manually.

    The problem in concrete terms

    A typical service business receives enquiries through

    • The website form, going to a shared address.
    • Direct email to whoever’s address the customer had.
    • Text to the owner’s mobile.
    • Facebook or Instagram messages.
    • Google Business Profile messages.
    • WhatsApp.
    • Phone, with voicemail somewhere else again.
    • The booking platform’s own messaging.

    No single view. No ownership. No record of who replied to what. The missed enquiry is not carelessness, it is a structural inevitability.

    What a unified inbox does

    Threads every channel by customer, so the text from Tuesday and the email from Friday sit in one conversation.

    Assigns ownership, so every conversation has a name against it.

    Records the history, so anybody picking it up has context.

    Measures response time, which is the metric that actually predicts conversion.

    Threading by customer, not by channel

    This is the feature that matters and not every platform does it well.

    Test during a trial: send yourself a message from three channels using the same phone number and email. Do they appear as one conversation or three?

    Matching keys

    • Phone number, normalised. This is why phone format standardisation matters.
    • Email address.
    • Social profile ID, linked manually the first time.

    Merge duplicates when they appear. A customer who texts from a second number will create a new record, and the merge should be one click.

    Assignment and ownership

    Every conversation gets an owner. No exceptions.

    Rules that work

    • Unassigned queue is checked by a named person at set times, not “whenever someone looks”.
    • First responder owns it until they hand it over explicitly.
    • Handover is a deliberate action, with a note, not a silent reassignment.
    • Nobody goes on holiday without reassigning their open conversations.

    Avoid round-robin auto-assignment in a small team. It sounds fair and it produces conversations owned by whoever happens to be up a ladder.

    Response time as the core metric

    First-response time is the number to manage.

    Set an internal target and publish it to the team

    Channel Target during hours
    Phone Answered, or called back within 30 minutes
    SMS and chat Within 15 minutes
    Social message Within 1 hour
    Email enquiry Within 1 hour
    Email, existing customer, non-urgent Same day

    Measure the median, not the average. One conversation left over a weekend distorts an average badly and hides the typical experience.

    An automatic acknowledgement is not a response. Count the first human reply. Acknowledgements are useful and they are not the metric.

    Automated tagging

    Tag automatically. Respond manually.

    Useful tags

    • Enquiry type. New job, existing customer, complaint, supplier, spam.
    • Urgency. Detected from keywords, with a bias toward over-flagging.
    • Service type, for routing to the right person.
    • Status. Awaiting us, awaiting customer, closed.

    The awaiting-customer tag is quietly valuable. It stops conversations sitting in the queue looking unanswered when the ball is with them, and it lets you chase properly after a few days.

    Do not auto-respond with anything substantive. An acknowledgement, yes. An answer, no.

    Templates without sounding templated

    Save the openings and the facts, not the whole message.

    Good snippet

    Our standard call-out is $95, which covers the first hour on site. Parts are quoted separately before any work goes ahead.

    Then write the rest personally. Snippets for the repeated factual content, human writing around it.

    Avoid full canned replies for anything beyond acknowledgement. They are recognisable and they read as dismissive on the receiving end.

    Out of hours

    • Set expectations in an auto-reply, with the emergency route.
    • Do not promise a response time you will not meet. Under-promise here.
    • Emergency path must be genuinely monitored. An emergency number nobody answers is worse than not offering one.
    • Turn off notifications for the team outside hours, deliberately. Always-on inboxes produce burnout and worse responses.

    Choosing a platform

    Check

    • Which channels are supported natively, especially Google Business Profile messages and WhatsApp, which are commonly missing.
    • Threading by customer, tested properly.
    • Mobile app quality. Most responses will be sent from a phone.
    • Does it connect to your CRM, so conversation history appears on the customer record.
    • Number portability. Can you keep your existing business number?
    • Export. Your conversation history is a business record.

    Many all-in-one platforms include this, which is usually the right answer at small scale rather than a separate tool.

    Measure it

    • Median first-response time, by channel, weekly.
    • Conversations with no owner, which should be zero at end of day.
    • Conversations open over 48 hours.
    • Enquiry to booking conversion, segmented by response time. This is where you prove the value of the whole exercise internally.
    • Messages received per channel, which frequently reveals a channel nobody was monitoring.

    Audit where messages currently arrive, listing every channel honestly. Most businesses find at least two nobody is responsible for, and that audit alone is worth doing before you buy anything.

    Need a pro to set it up? [BOOK A CALL]

  • Automating Time Tracking and Payroll Processing

    Automating Time Tracking and Payroll Processing

    Paper timesheets are wrong in both directions. People forget hours they worked and estimate hours they did not, and neither error is deliberate.

    TL;DR Track time against jobs, not just against the day. Geofencing should prompt, not clock automatically. Be transparent with the team about what is tracked. Sync to payroll rather than retyping it.

    Job-costed time, not just attendance

    Attendance tracking tells you somebody worked eight hours. Job-costed time tells you which jobs are profitable.

    Capture

    • Clock in and out for the day.
    • Time per job, start and finish.
    • Travel time, separately from job time.
    • Breaks.
    • Non-billable time, categorised. Merchant runs, admin, training, vehicle.

    The non-billable categories are where the insight is. A technician at 60 percent billable is not slacking. Something structural is consuming their day, and the categories tell you what.

    This is the number that tells you if your pricing is right. Quoted three hours, actual five, across a whole job type, means the price is wrong and you would never see it from a day-total timesheet.

    Geofencing, done properly

    Use it to prompt, not to clock automatically.

    Automatic clocking on location entry causes

    • Clock-ins when driving past a site.
    • Failures when GPS drifts, which it does.
    • Clock-outs during a lunch break off site.
    • Genuine resentment, because it feels like being monitored rather than supported.

    The prompt version

    You have arrived at 14 Elm Road. Start the job? [Yes] [Not yet]

    One tap, accurate, and the person remains in control. Compliance is dramatically better because it feels like a tool rather than a tracker.

    Battery matters. Continuous high-accuracy GPS drains phones. Check what the app does in the background and test it over a full working day before rolling it out.

    Be transparent with the team

    Introduce it as accurate pay and better quoting, because that is what it should be.

    Say clearly

    • What is tracked. Location at clock events, time per job.
    • What is not. Location outside working hours, personal phone activity.
    • Who sees it. By role.
    • How long it is kept.
    • What it is used for. Payroll accuracy and job costing.
    • What it is not used for. Policing break times.

    Put it in writing. In many jurisdictions employee monitoring carries legal notification and data protection obligations. Check what applies before you switch anything on.

    Then honour it. The first time somebody is challenged over a fifteen-minute discrepancy, the system becomes an adversary and the data quality collapses.

    The upside for the team

    Frame it honestly, because the benefits are real.

    • Overtime captured automatically. Most under-reporting is in the team’s disfavour.
    • No Sunday evening timesheet.
    • Disputes resolved by record, not by memory.
    • Travel time counted properly, which paper timesheets routinely lose.

    Lead with the overtime point. It is true and it reframes the whole thing.

    Overtime calculation

    Automate it, because manual calculation is where payroll errors live.

    Configure

    • Standard hours and the overtime threshold.
    • Rates by band.
    • Weekend and holiday rules.
    • On-call and call-out payments.
    • Rounding rules, stated explicitly.

    Rules vary by jurisdiction and can be strict, particularly on overtime thresholds, rounding practices and record retention. Get the configuration checked against the rules that apply to you rather than assuming the software defaults are compliant.

    Have somebody approve the period before it reaches payroll. Automated calculation with human approval, not automated payment.

    Payroll sync

    The goal is nobody retyping numbers.

    Flow

    1. Time captured in the field app.
    2. Manager reviews and approves, with exceptions flagged.
    3. Approved hours push to payroll.
    4. Payroll processes.
    5. Costs post back to job records.

    Step five is the one people skip and it is what makes job costing real.

    Check the integration before buying. Native connection to your payroll provider, not a CSV export you reformat every fortnight.

    Exceptions to watch

    Build alerts for these rather than reviewing everything.

    • Missing clock-out. Somebody forgot. Common and easily fixed if flagged same day.
    • Unusually long job time, against the type average.
    • Clock event outside a geofence.
    • Overtime above a threshold, which may indicate a scheduling problem rather than extra work.
    • Zero non-billable time, which is not realistic and means categories are not being used.

    Review exceptions, approve the rest. Reviewing every line defeats the purpose.

    Measure it

    • Billable percentage per technician and overall.
    • Quoted versus actual hours by job type. The most commercially useful output.
    • Payroll processing time, before and after.
    • Payroll corrections per period. Should fall to near zero.
    • Non-billable time by category, which shows you where the day is going.

    Start by tracking time against jobs rather than just against the day. Even without geofencing, quoted-versus-actual by job type will show you within a month which of your services is quietly unprofitable.

    Need a pro to set it up? [BOOK A CALL]

  • Implementing Zapier and Make for No-Code Automation

    Implementing Zapier and Make for No-Code Automation

    No-code automation is genuinely powerful and quietly fragile. Build it properly or it fails silently for three weeks before anybody notices.

    TL;DR Prefer native integrations over bridges. Build error notification before you build anything else. Start with lead capture and review requests. One automation at a time, tested with real data.

    Native first, always

    Before building a bridge, check whether the two systems already talk to each other.

    Why native wins

    • Fewer moving parts.
    • No per-task cost.
    • Maintained by the vendor.
    • Faster, and no queueing delay.
    • Does not break when a plan limit is hit.

    Use a bridge when there is no native option, or the native one does not do the specific thing you need.

    Every bridge is a permanent maintenance obligation. Count them, and be reluctant to add the tenth.

    The five worth building first

    1. Web form to CRM to instant notification

    Enquiry submitted, contact created, alert to whoever responds, auto-acknowledgement to the customer.

    Thanks for getting in touch. We have got this and will come back to you within the hour during working hours. If it is urgent, ring 01234 567890.

    This one automation addresses the biggest cause of lost enquiries, which is delay.

    2. Job completed to review request

    Job marked complete, wait 24 hours, send the review request, but only if no complaint flag is set.

    The conditional check is essential. Sending a review request to someone with an open complaint is a self-inflicted wound.

    3. Missed call to text back

    Missed call detected, automatic text within a minute.

    Sorry we missed you, this is Sicc Heating. We are on a job. Reply here with what you need and we will come straight back.

    Cheap to build and it recovers callers who would otherwise ring the next number.

    4. Payment received to thank you and next step

    Payment lands, receipt sent, thank-you message, next service pencilled or reminder scheduled.

    5. Quote sent to follow-up sequence

    Quote issued, four-touch sequence begins, stops on any reply or acceptance.

    The stop condition is the part people forget and it is what prevents automated chases arriving mid-conversation.

    Error handling, built first

    This is the section everybody skips and then regrets.

    Every automation needs

    • A failure notification. Email or text to a person, not a log nobody reads.
    • A weekly check. Two minutes looking at the run history.
    • A manual fallback. What happens if it is down for a day? Somebody should know.
    • Retry logic for transient failures, with a limit.

    Common silent failures

    Failure Symptom
    Plan task limit hit Everything stops at once, mid-month
    API credentials expired One integration stops, others fine
    Field renamed in the source system Runs successfully, writes blanks
    Rate limit Intermittent, hardest to spot
    Vendor changed their API Sudden, total, for one connection

    The blank-field failure is the worst, because the automation reports success while writing nothing useful. Add a validation step that checks required fields are populated and alerts if not.

    Building safely

    One at a time.

    1. Write down what it should do, in a sentence, before opening the tool.
    2. Build it.
    3. Test with real data, in a test record.
    4. Run it manually a few times, checking the output each time.
    5. Turn it on for a small segment, if possible.
    6. Watch it for a week.
    7. Then leave it alone.

    Document each one. What it does, what triggers it, what breaks if it stops, who built it. In six months you will not remember, and if somebody else needs to fix it they will have nothing to work from.

    Multi-step workflows

    Longer chains multiply the failure modes.

    Rules

    • Filter early. Drop irrelevant records at step one, not step six. Cheaper and faster.
    • Use paths rather than separate automations where the logic is genuinely branching.
    • Add a formatting step rather than assuming data is clean. Phone numbers and dates in particular.
    • Keep it under about eight steps. Longer chains are hard to debug and hard to reason about.
    • Never chain two bridges together. If automation B is triggered by automation A, you have doubled the failure surface for no benefit.

    Cost control

    Task-based pricing surprises people at volume.

    • Filter before the action, so you are not paying for tasks that do nothing.
    • Batch where timing allows. A daily digest instead of fifty individual runs.
    • Check the run history for waste. Automations firing on records that get filtered out later are pure cost.
    • Watch the trend as you grow. The plan that was fine at fifty jobs a week is not at two hundred.

    What not to automate

    • Complaints. Always a human.
    • Anything sending money.
    • Deleting records. Archive instead.
    • Anything requiring judgement about a customer relationship.
    • Compliance-critical steps where a silent failure means a legal problem.

    Rule: if a silent failure would be expensive, do not automate it without a check that it ran.

    Measure it

    • Time saved, estimated honestly. Tasks per week times minutes per task.
    • Failure rate per automation, monthly.
    • Cost per month, tracked against the plan limit.
    • Automations you no longer understand. If any exist, document or delete them.

    Build the missed-call text-back this week, with a failure notification attached. It is a single trigger and a single action, it takes twenty minutes, and it recovers leads you are currently losing to a voicemail box.

    Need a pro to build it? [BOOK A CALL]

  • Automating Inventory and Material Replenishment

    Automating Inventory and Material Replenishment

    The mid-job trip to the merchant costs an hour of unbilled time, delays the customer, and happens because nobody counted the stock.

    TL;DR Track the twenty parts that cover most jobs, not everything. Set minimum thresholds and reorder automatically at those points. Scan materials to the job so your costing is real. Van stock is where the time is won.

    Track the twenty, not the two hundred

    Full inventory management is the wrong goal for a small service business. The effort of tracking every washer exceeds the value.

    Do a Pareto pass

    1. Pull twelve months of parts usage.
    2. Sort by frequency, not by cost.
    3. The top twenty or so items will cover the large majority of jobs.

    Automate those. Manage the rest by exception.

    The high-frequency items are what cause mid-job trips. A rare specialist part is ordered per job anyway and was never going to be on the van.

    Van stock

    This is where the returns are.

    Build a van stock list per technician or per van type

    • The high-frequency items, in the quantity that covers a typical week.
    • Consumables, generously. They are cheap and running out is expensive.
    • Anything needed for the most common emergency call.

    Standardise the layout. Same part, same place, every van. It saves time and makes stock checks visual rather than forensic.

    Weekly restock at a fixed time. One person, one trip, all vans. Compare that against each technician making their own trips as they run out.

    Track mid-job supplier trips as a metric. Every one is a signal about what should be on the van, and the pattern becomes obvious within a month.

    Minimum stock thresholds

    For each tracked item, set a reorder point and a reorder quantity.

    Reorder point should cover usage during the supplier lead time, plus a buffer for a bad week.

    Item: 15mm compression elbow
    Weekly usage: ~40
    Lead time: 2 days
    Reorder point: 60
    Reorder quantity: 200

    Review the thresholds quarterly. Seasonal work changes usage substantially and static thresholds cause stockouts in exactly the busiest months.

    Automatic purchase orders

    Threshold hit generates the order. A human approves it.

    The flow

    1. Stock falls to the reorder point.
    2. Draft PO generated for the preferred supplier.
    3. Alert to whoever approves.
    4. One-tap approve, sends to the supplier.
    5. Goods in updates the stock level.

    Keep the approval step. Fully automatic ordering produces duplicate orders, orders during a supplier price dispute, and orders for items you are discontinuing.

    Set a value threshold below which it auto-approves. Small consumable orders do not need a decision.

    Scanning materials to jobs

    This is what makes job costing real rather than estimated.

    How

    • Barcode or QR scan on the part, in the field app, at point of use.
    • Assign to the job automatically.
    • Stock decrements, job cost increments.

    What it unlocks

    • True margin per job, not estimated.
    • Accurate quoting, because you know real consumption rather than what you assumed.
    • Stock accuracy without counting.
    • Shrinkage visibility. Uncomfortable, and worth knowing.

    Adoption is the hard part. If scanning adds friction, technicians will not do it, and partial data is worse than none. Test the flow on a real job before rolling it out and count the taps.

    A workable lower-tech version: select from a short list of common parts instead of scanning. Less accurate, far higher compliance.

    Supplier relationships

    Automation does not replace this.

    • Consolidate suppliers where possible. Better pricing, fewer accounts, simpler ordering.
    • Negotiate on your actual volume, which you now know precisely.
    • Ask about stock holding for your high-frequency items. Many will hold stock for a regular account.
    • Delivery to site, which removes the collection trip entirely.
    • Keep a second supplier warm for the items you cannot work without.

    The data from tracking is your negotiating position. “We buy 8,000 of these a year” is a different conversation from “we buy a few”.

    Waste and shrinkage

    • Reconcile periodically. Physical count against system, on the tracked items only.
    • Investigate patterns, not incidents. One discrepancy is noise. A consistent one is a process problem or a person.
    • Check for over-ordering, particularly of things that expire or supersede.
    • Return unused job-specific parts promptly, while they are still returnable.

    Tooling

    Most field service platforms include enough inventory function for a business at this scale. Use that before buying anything separate.

    What matters

    • Stock levels visible in the field app.
    • Parts assignable to jobs quickly.
    • Reorder alerts.
    • PO generation.
    • Per-van stock, not just a single warehouse figure.

    Per-van tracking is the feature most often missing and the one that matters most for a mobile business. Check for it specifically.

    Measure it

    • Mid-job supplier trips per week. The headline number.
    • First-time fix rate. Rises directly with correct van stock.
    • Stock value held, which should not grow faster than revenue.
    • Stockouts on tracked items. Should approach zero.
    • Material cost as a percentage of job value, tracked over time. Drift means quoting is out of date.

    Pull twelve months of parts usage this week and identify your top twenty items by frequency. Stock those properly in every van and the mid-job merchant run largely disappears.

    Need a pro to automate it? [BOOK A CALL]

  • Building a Video Training Academy for Staff

    Building a Video Training Academy for Staff

    Every time you explain the same thing to a new hire, you are paying twice for something you could have recorded once.

    TL;DR Record in the moment rather than scheduling a studio day. Two to five minutes per video, one topic each. Name them so they are findable. Add a short check so you know it was watched and understood.

    Record in the moment

    The reason training libraries never get built is that they get scheduled.

    The method that works

    Next time you explain something to somebody, hit record. Screen record for software, phone camera for physical tasks. Explain it exactly as you would anyway. Stop recording. Upload. Name it.

    Total additional time: about ninety seconds.

    Twenty videos in a month from work you were doing regardless. That is a library.

    Do not re-record for polish. Nobody watching an internal training video cares about production. They care whether it answers the question.

    Length and scope

    Two to five minutes. One topic.

    Why short wins

    • Findable. Somebody needing one answer does not want a forty-minute induction video.
    • Rewatchable on site, on a phone, standing in a hallway.
    • Cheap to replace when the process changes.
    • Actually watched. Long videos get skipped and then claimed as watched.

    Split anything longer. “Using the job app” becomes eight videos: opening a job, adding photos, recording parts, completing, and so on.

    What to record first

    Priority Type Example
    1 Asked repeatedly “How do I add a part to a job?”
    2 Week-one essentials Van setup, app basics, site arrival
    3 Safety and compliance Isolation procedures, PPE, certification steps
    4 Only one person knows The single-point-of-failure knowledge
    5 Quality standards What good looks like, with examples

    Start with the questions you answered this week. That list writes itself and it is the highest-value content.

    Naming and organisation

    Findability is the entire difference between a library and a folder of videos.

    Naming convention

    [Area] - [Task] - [Version date]

    App - Adding parts to a job - 2026-03
    Site - Arrival and customer greeting - 2026-01
    Safety - Gas isolation procedure - 2025-11

    Structure

    • Group by when it is needed, not by department. “Week one”, “On site”, “In the app”, “Safety”.
    • A single index page listing everything with links.
    • Search that works. If your platform’s search is poor, the index page carries the load.
    • Link videos from the point of use. A QR code on the equipment beating any portal.

    Knowledge checks

    Watching is not learning, and you need to know which happened.

    Keep checks minimal

    • Three questions per video, maximum.
    • Practical, not recall. “A customer says the pressure gauge is at zero. What do you check first?” beats “what is the correct operating pressure range?”
    • Open book. The point is competence, not memory.
    • Retake freely. This is training, not an exam.

    For physical skills, a check is a supervised task, not a quiz. Record who signed it off and when.

    Track completion, because unwatched safety training is a liability as well as a gap.

    Onboarding structure

    Sequence the library into a week-one path.

    Day 1: Who we are, how we work, safety essentials, app setup. Four videos, plus a real conversation.
    Day 2: Site arrival, customer greeting, the standard job walkthrough. Then shadowing.
    Day 3: Job completion, photos, records, invoicing. Shadowing.
    Day 4: Common jobs, one video each. Supervised work.
    Day 5: Review, questions, first solo job with check-in.

    Video replaces the repeated explanation, not the human contact. A new hire watching videos alone all week will leave. Pair every module with real supervised work.

    Standardising delivery

    The commercial argument for all of this.

    Record the standard

    • What a completed job looks like, photographed properly.
    • The exact words for the customer walkthrough.
    • How the site should be left.
    • The pencil-it question for the next service.

    Show good and bad side by side. Two photos of the same finished job, one meeting standard and one not, teaches more than a paragraph of description.

    Consistency is what lets you scale. Customers accept a different technician if the experience is identical. They do not if it is a lottery.

    Keeping it current

    Out-of-date training is actively harmful.

    • Date every video visibly.
    • Review annually, and immediately when a process or interface changes.
    • Archive rather than delete, so nobody follows an old version by accident.
    • Anybody can flag a wrong video in one message, and gets thanked for it.

    Re-recording is cheap. That is the advantage of the informal format. A polished production nobody wants to redo becomes permanently out of date.

    Tooling

    Whatever you already have.

    • Screen recording with a shareable link, for software.
    • Phone camera for physical tasks.
    • A shared drive or an existing team platform for storage.
    • A simple document as the index.

    Do not buy a learning management system for a team of ten. The overhead exceeds the benefit and it becomes another login nobody uses.

    Measure it

    • Time to first solo job for a new hire.
    • Repeat questions. The clearest signal of a missing video.
    • Completion rates on the week-one path.
    • Consistency of job outcomes across technicians. Callbacks by person is the honest measure.

    Record one video tomorrow, the next time somebody asks you a question you have answered before. Ninety seconds of extra effort, and you never explain it a third time.

    Need a pro to build it? [BOOK A CALL]

  • Automating Fleet and Route Optimization

    Automating Fleet and Route Optimization

    Drive time is the largest unbilled cost in a field service business and the one nobody measures.

    TL;DR Cluster jobs geographically before optimising routes. Send an automatic on-the-way text, which removes most inbound calls. Measure drive time as a percentage of the working day. Tracking is for dispatch, not for surveillance.

    Measure drive time first

    Most operators are surprised by their own number.

    Calculate it. Total driving hours divided by total working hours, per technician, over a fortnight.

    Then split it

    • Between jobs. The bit optimisation actually improves.
    • To the first job and home from the last. Improved by area assignment, not by routing.
    • Supplier and depot trips. Improved by van stock, covered separately.

    Knowing which of the three dominates tells you which fix to invest in. People buy routing software when their real problem is three trips to the merchant a day.

    Cluster before you optimise

    Geographic clustering delivers more than clever routing does.

    How it works

    • Divide your area into zones.
    • Assign days, or technicians, to zones.
    • Book non-urgent work into the zone day.

    We are in your area on Tuesdays and Thursdays. Which suits better?

    Customers accept this readily. It sounds organised, which it is.

    What it produces

    • Much shorter hops between jobs.
    • Predictable days for the team.
    • Genuine capacity to absorb an emergency without wrecking the schedule.
    • Better first-time fix rates, because the van stock can match the zone.

    Keep emergency capacity outside the system. A rigid zone schedule with no slack breaks the first time something urgent lands.

    Route optimisation software

    Useful once clustering is in place, and much less useful before.

    What to look for

    • Time windows, not just addresses. A route that ignores appointment times is useless.
    • Job duration estimates by type.
    • Technician skill matching, so the right person gets the right job.
    • Live re-optimisation when something overruns.
    • Real traffic data.
    • Return-to-base handling, if relevant.

    Realistic expectation. Routing software delivers useful gains on top of good clustering. It does not rescue a badly organised schedule, and vendor case study percentages usually describe businesses starting from chaos.

    The on-the-way text

    The single highest-value automation in this whole area, and the cheapest.

    Two messages

    Night before: Sicc Heating. We are booked in tomorrow between 8 and 10am. Reply if you need to change it.

    On the way: On our way, about 20 minutes. Dave is driving, he will call if he cannot find parking.

    What this eliminates

    • The “what time are you coming” call, which is the highest-volume inbound call in most service businesses.
    • Missed appointments through nobody being home.
    • The complaint about waiting in all day.

    Automate both from the schedule. The night-before from the calendar, the on-the-way from the technician marking travel started in the app.

    Naming the technician matters more than it should. It converts an unknown van into a person.

    Live tracking

    Frame it as dispatch and customer service, not monitoring.

    Legitimate uses

    • Deciding who takes an incoming emergency.
    • Giving an accurate arrival estimate.
    • Confirming attendance for a dispute.
    • Safety, for lone workers.

    Where it goes wrong

    • Tracking outside working hours.
    • Using it to challenge break times.
    • Introducing it without telling the team.

    Be explicit with staff about what is tracked, when, and who sees it. In many jurisdictions this is a legal requirement as well as the right thing to do. Covert tracking of employees is a serious problem, not a management technique.

    Customer-facing tracking is now expected on the day of the appointment. A live map link on the on-the-way text is a genuine differentiator for smaller operators.

    Reducing supplier trips

    Often the biggest single drive-time saving, and it has nothing to do with routing.

    • Van stock lists by job type. The twenty parts that cover most jobs.
    • Weekly restock at a fixed time, not ad hoc.
    • Delivery to site for large or job-specific items.
    • Consolidated collection by one person for the whole team.

    Track mid-job supplier trips. Each one is roughly an hour of unbilled time plus a delayed job, and the pattern in them tells you exactly what to add to the van.

    Fuel and vehicle costs

    • Fuel cards for data, not just payment. The reporting is the point.
    • Cost per job including fuel and vehicle time, not just fuel per mile.
    • Idling, which is invisible without telematics and adds up.
    • Servicing on schedule. A vehicle off the road unexpectedly costs far more than the service.

    Measure it

    • Drive time as a percentage of working hours. The headline metric.
    • Jobs completed per technician per day.
    • Miles per job.
    • First-time fix rate. Rises with better van stock and zoning.
    • “Where are you” calls per week. Should fall to near zero with the on-the-way text.

    Set up the night-before and on-the-way texts this week. Two automations, both trivial to build, and together they remove the most common inbound call in your business.

    Need a pro to set it up? [BOOK A CALL]

  • AI-Powered Phone Receptionists and Call Triage

    AI-Powered Phone Receptionists and Call Triage

    A missed call in a service business is usually a lost job, because the caller rings the next number on the list within ninety seconds.

    TL;DR Disclose that it is automated. Route emergencies to a human immediately, never to a bot. Capture and confirm rather than trying to fully resolve. Test it by calling it yourself, angrily.

    Disclose it, always

    Start here because it is both a trust issue and, increasingly, a legal one.

    In the US, the FCC treats AI-generated voices as artificial or prerecorded for the purposes of the TCPA, which brings consent requirements, particularly for outbound calls. Rules differ by jurisdiction and are moving. Check what applies where you operate.

    Practical position

    • Say it is automated in the opening line. “Hi, this is the automated assistant for Sicc Heating.”
    • Never impersonate a named person.
    • Offer a human immediately. “Say ‘person’ at any point and I will put you through.”
    • Inbound is far less fraught than outbound. Someone who called you is a different consent situation from someone you called.

    People mind an automated system far less than they mind being deceived by one.

    What it should actually do

    Capture and route. Not resolve.

    Good use

    • Take the name, number, address and nature of the problem.
    • Classify urgency.
    • Confirm what it heard, back to the caller.
    • Route or escalate.
    • Text a confirmation to the caller.
    • Log the whole thing with a transcript.

    Bad use

    • Quoting prices.
    • Diagnosing faults.
    • Handling complaints.
    • Negotiating anything.
    • Making promises about timing.

    The test: could this response be wrong in a way that costs money or trust? If yes, a human handles it.

    Emergency routing

    This is where a badly configured system does real damage.

    Build an explicit emergency path

    • Keyword and phrase detection. Leak, flood, no heat, smell of gas, sparking, no power, elderly, vulnerable, baby.
    • On detection, stop the script. Do not continue collecting structured information.
    • Transfer to a live number immediately, or take the number and trigger an urgent alert.
    • If nobody answers, say so honestly and give the emergency alternative.

    That sounds urgent. I am putting you through to someone now.

    Default to escalating. A non-emergency escalated costs you a phone call. An emergency handled as routine costs you far more than that.

    Anything gas-related or electrical-safety-related should escalate unconditionally.

    Call routing rules

    Beyond emergencies, keep the tree shallow.

    Caller type Route
    Emergency Human, immediately
    Existing customer with an issue Human during hours, urgent callback outside
    New enquiry Capture full details, book if possible, confirm by text
    Existing customer, routine booking Book directly, confirm by text
    Supplier or admin Message, normal callback
    Sales calls Politely ended

    Three options maximum at any prompt. Long menus are why people hate phone systems, and the assistant does not need them if it can understand plain speech.

    Booking through voice

    Works well for routine, standardised work. Poorly for anything requiring assessment.

    If you enable booking

    • Confirm every detail back before writing it in.
    • Send a text confirmation immediately. Voice-only confirmation gets misheard and misremembered.
    • Hold rather than commit for anything unusual, then have a human confirm.
    • Do not let it book emergencies. Those need a person deciding.

    Out of hours

    This is where the return is clearest.

    Overnight and weekend calls fall into three groups

    1. Genuine emergencies. Escalate to the on-call number.
    2. Non-urgent, will book in the morning. Capture and confirm. This is the majority.
    3. Wrong number and sales calls. Filter.

    Group two is the win. Those callers currently ring you, get voicemail, and ring somebody else. Capturing them properly with a same-morning callback promise recovers a meaningful share.

    Test it properly

    Call it yourself, repeatedly, in the ways real customers call.

    • Calmly, with a straightforward request.
    • Angrily, mid-complaint. It should recognise this and escalate.
    • Panicked, with an emergency. Time how long to a human.
    • With background noise. A road, a busy house.
    • With a strong local accent and local place names. Address recognition is where these systems most often fail.
    • Interrupting it.
    • Saying something completely off-script.

    Then have somebody unfamiliar with it call. You know what it expects. Customers do not.

    Review transcripts weekly for the first month. You will find failure patterns you did not anticipate, particularly around street names.

    The honest limitations

    • Accents and noise degrade accuracy meaningfully.
    • It cannot read tone reliably. Frustration often reads as neutral.
    • Local knowledge is absent unless you supply it explicitly.
    • It will occasionally invent something. Constrain it to reading from your data rather than generating.
    • Some customers will hang up, particularly older ones. That is a real cost against the calls you save.

    Compare it against your actual current state, which is voicemail, not against a perfect human receptionist.

    Measure it

    • Missed calls before and after. The headline.
    • Calls handled entirely without a human.
    • Escalation rate, and whether escalations were appropriate.
    • Booking rate from automated capture versus human answer.
    • Hang-up rate in the first fifteen seconds. Rising means the opening is wrong.
    • Emergencies missed. Must be zero. Audit this specifically.

    Set it up for out-of-hours only to begin with, then call it yourself six times in six different moods before it takes a single real call.

    Need a pro to configure it? [BOOK A CALL]