The systems we build,
drawn to scale

Most agencies show you a wall of logos you can’t verify. We publish the drawings instead — what each system replaces, how it works end to end, and where a human stays in the loop. Judge the mechanics, not the name-dropping.

THESE ARE REFERENCE BUILDS, NOT CLIENT CASE STUDIES. NO CLIENT IS DESCRIBED OR IMPLIED ON THIS PAGE. THE MECHANICS ARE REAL; EVERY FIGURE IS A WORKED EXAMPLE, NOT A MEASURED RESULT.

The intake desk that never closes

REFERENCE BUILD — NOT A CLIENT CASE STUDY

The mess

Enquiries land in a shared inbox that four people half-own, so each assumes another has picked it up. The good ones get answered late; the ambiguous ones get read, left, and quietly forgotten. Nobody can say how many came in last week or where any of them ended up, because the only record is a thread somebody starred.

The system

  1. T-01

    An enquiry arrives — shared inbox, web form or a note from a call — and is acknowledged immediately, so nobody is left wondering whether it landed.

  2. T-02

    An AI triage step reads it against your real service descriptions and policies, classifies what it is, and asks the qualifying questions a good coordinator would.

  3. T-03

    The record is written to the CRM, routed to the right person by service line or workload, and — where it’s a booking — put straight into the calendar.

  4. T-04

    Anything odd — an unhappy client, an unusual request, low confidence — routes to a human with the full thread attached. Judgment stays with people.

STACK: AI triage · CRM · shared inbox · booking

Schematic: emails, web forms and call notes flow into an AI triage step, which writes qualified enquiries into the CRM and calendar or hands off to the team with full context. CALL NOTE EMAIL / FORM AI TRIAGE STEP READS YOUR REAL DOCUMENTS CRM + BOOKING TEAM HANDOFF ACKNOWLEDGED ON ARRIVAL · ROUTED IN SECONDS
FIG. 01 DESIGNED OUTCOME: Every enquiry acknowledged, qualified and routed within seconds of arriving

The math

ILLUSTRATIVE
Hours reclaimed ≈ 9 hrs/week of enquiry triage, chasing and re-typing
Loaded cost £28/hour
Yearly value 9 × £28 × 52 ≈ £13,000
Payback Worked out against your real numbers in the blueprint, before you commit

The invoice chaser that never gets awkward

REFERENCE BUILD — NOT A CLIENT CASE STUDY

The mess

The work is done, the invoice is out, and now someone has to write “just checking in on this” to a client they want to keep — without sounding desperate. So it gets put off. Six hours a week go into drafting, softening and re-softening follow-ups, and the invoices that slip past thirty days are exactly the ones nobody wants to touch. Cash flow runs on whoever feels brave that week.

The system

  1. T-01

    The accounting system flags an invoice past its due date; the workflow wakes on a schedule, not when someone remembers.

  2. T-02

    AI drafts a follow-up in your voice — referencing the project, the amount, and how overdue it is. The tone escalates one polite notch per round.

  3. T-03

    A human approves or edits every draft in one click before it sends. Nothing reaches a client unread. Each send is logged against the invoice.

  4. T-04

    The cycle repeats on schedule until payment lands. Anything past round three gets flagged for a personal call — some conversations should stay human.

STACK: Accounting API · scheduled workflow · AI drafting · approval step

Schematic: overdue invoices flow from the ledger to an AI draft, then a one-click human approval, then a logged send — repeating on schedule until the invoice is paid. LEDGER OVERDUE FLAG AI DRAFT IN YOUR VOICE APPROVE HUMAN GATE SEND + LOG REPEATS ON SCHEDULE UNTIL PAID CHASE TIME: HOURS A WEEK — NOW MINUTES OF APPROVALS
FIG. 02 DESIGNED OUTCOME: Receivables chased politely and relentlessly; relationships intact

The math

ILLUSTRATIVE
Hours reclaimed ≈ 6 hrs/week of drafting and re-softening follow-ups
Loaded cost £35/hour
Yearly value 6 × £35 × 52 ≈ £11,000
Payback Before faster payment does anything at all for cash flow

The spreadsheet that became a system

REFERENCE BUILD — NOT A CLIENT CASE STUDY

The mess

Every quote runs through one spreadsheet — fourteen tabs of volume breaks, client tiers and rate cards — and one person who actually understands it. When they’re out, quotes wait. When someone guesses, the quote goes out below margin and nobody notices until month-end. The pricing brain of the business lives in a file called FINAL_v9.

The system

  1. T-01

    Someone opens the quoting app and picks the client and line items — no spreadsheet, no hunting through fourteen tabs.

  2. T-02

    The tribal knowledge from the old sheet — volume breaks, client tiers, rate cards — runs as code against live cost data in Postgres.

  3. T-03

    Any line priced below the margin floor is blocked at entry. The override exists, but it requires a manager’s sign-off — the escalation is a person.

  4. T-04

    The finished quote renders to PDF and logs to the CRM, and a scheduled workflow nudges the owner if the client goes quiet. The keeper of the sheet gets their week back.

STACK: Internal quoting app · rules engine · database

Schematic: the team works in a quoting app backed by margin rules in Postgres; quotes render to PDF and log to the CRM, and below-margin lines are blocked at entry. YOUR TEAM QUOTING APP MARGIN GUARD QUOTE PDF CRM LOG QUOTE OUT: MINUTES — BELOW MARGIN: BLOCKED
FIG. 03 DESIGNED OUTCOME: Accurate quotes in minutes, with margins protected by rules

The math

ILLUSTRATIVE
Hours reclaimed ≈ 12 hrs/week across the team and the keeper of the sheet
Loaded cost £30/hour
Yearly value 12 × £30 × 52 ≈ £18,700
Payback Before counting the below-margin quotes that now never go out

Want something like one of these?
Start with the service it belongs to

Each of these is a configuration of the two things we build. Yours will look different — they always do. The free review tells you which shape fits, and whether it’s worth building at all.

Next step

Stop doing work a system should do.

Twenty minutes on the process that is costing you most, and a straight answer on whether it is worth building. If it isn’t, we’ll say so.

Book a free automation review

20–30 minutes · No pitch · We’ll tell you straight if it isn’t worth building