Confidential · Prepared after our call on 4 June

ES Group × AI Operational Audit — what we discussed

A summary of the call so you and your brother can review before Friday’s follow-up, plus the audits closest in shape to what ES Group is running.

Devon Kay · ES Group · 25–30K jobs/yr · 41-min discovery · 4 Jun 2026

The picture from the call

Captured from your own words so the audit picks up exactly where the call left off.

  • The headline pain is double-keying. Field staff enter job info into ES Group’s ERP, then admin re-enters that same data into each client’s separate ERP. No API bridge possible across ~50 different client systems. This is the biggest weight-off-the-shoulders win in the business.
  • It’s a high-volume operation. 25,000–30,000 jobs/yr, jobs ranging $100 to $600–700K. Same number of admin touchpoints whether the job’s worth $150K or $700K — so admin load scales linearly with job count, not revenue.
  • The 80/20 is clear. Your own words: “You take your 20% of your biggest clients… that would reduce most of the administrative work and then enable us to grow the administrative base.” The audit picks the top 5 clients as the first targets.
  • You want to grow efficiently — not by adding admin headcount. “Enabling the staff to do more work in the space where they have to, rather than having to add resources as we grow.”
  • Translating technician notes into client invoice descriptions. “Collating 20 days’ worth of work into five sentences” — classic AI-summarisation use case, low risk, high ROI.
  • The ERP’s own AI hasn’t hit the mark. Common pattern: built-in AI bolted onto disjointed tools doesn’t solve the connection problem. That’s the gap.
  • Lead gen / marketing is off the table. Relationship-based, person-to-person. We won’t scope work there.

Where we’d find the time

Real numbers come from the interviews — but based on what you described, these are the obvious places to dig.

  • Eliminating the double-key for the top 5 clients. AI reads what the technician filed in your ERP and writes it directly into the client’s system — in whatever format that client expects. 50 different client systems become 50 different output templates, not 50 admin re-entries. Single biggest hours-back item.
  • Technician-notes-to-client-invoice summarisation. AI takes raw technician notes (text or photo) and produces the polished sentence-per-day breakdown your client invoices need. Human reviews and approves — same as your current process minus the writing.
  • Smart photo / document classification. The change scenario you raised (client now needs photos as well as H&S docs) is handled the same way: AI tags and routes the photo to the right client field, you spot-check.
  • Where AI should NOT touch. The technician relationship, the client relationship, the safety-critical compliance work. The audit will tell you exactly which bits to keep human.

Three steps, in order

No big build up front. The audit comes first, then you decide what to implement — one piece at a time.

STEP 1

The audit

Structured interviews with department leads — admin, field-tech, accounts. We map the full operation and find where AI saves time, where it’s risky, and where it doesn’t belong.

STEP 2

The priority matrix

Every opportunity ranked by impact vs effort. You see exactly what’s cheap to build and high-impact — the quick wins to start with.

STEP 3

Phased build

We build the highest-value, lowest-cost item first. It works, you measure the return, you decide what’s next. Never one big bill.

Three layers. You can’t skip one.

Most failed AI projects skip straight to the top before the two underneath are ready. The audit makes sure that doesn’t happen at ES Group.

1

Foundation — is your data AI-ready?

Job records, technician notes, client account details, invoice history. We check the data state and clean what needs cleaning before AI touches it. Your own words: “rubbish in, rubbish out.” Exactly.

2

Systems — can your tools connect?

Your ERP has an API. The 50 client ERPs are the harder question. We confirm what talks to what — and where it doesn’t, AI fills the gap by reading and writing on a human’s behalf, no integration required.

3

Intelligence — where AI actually goes in

Only once 1 and 2 are sound do the AI co-worker plugins go in — double-key elimination, note-to-invoice summarisation, smart document routing.

Addressing the NZ tax risk you raised

No black-hole spending. Built for the way you’d want to be invested in.

“If we invest $100,000 into a piece of software and we don’t use it or it doesn’t work, it’s an additional $100,000 on tax at the end of the year… black-hole spending, basically.”— Devon, on the call
DE-RISK 1

Audit before any build

You don’t commit a cent to a build until the audit shows it works in your business and what it’s worth. The blueprint comes first — the build is your call.

DE-RISK 2

Money-back guarantee on the audit

If the audit doesn’t find a multiple of its own cost in recoverable waste, you get a full refund and keep everything we produced. Specifics on Friday.

DE-RISK 3

Piece by piece, never one big bill

Start with a quick win at a small build cost. See it working, measure the return, then decide what’s next. No $100K commitment that could become a tax problem if it doesn’t land.

A real client, on camera

Different industry from yours, but the same audit methodology — and Tom is the only client to date who’s come on camera by name. Useful as proof the process is real and people are happy with it.

Real client · named testimonial

“Above and beyond”

Tom Fay · Founder, T5 Football · ~600 active clients, $500–600K revenue
Tom called the work “above and beyond.” We found gaps his previous partners had missed entirely, mapped the full picture, and built a plan to move fast — delivered in a fraction of the expected timeline.
What the audit found: ~30 hours a week of manual admin on a 6-person operation. Manual reconciliation across multiple tools. Bank transfer names that couldn’t be matched to clients. A founder buried in admin instead of running the business.
~1,560 hrs/yr
Recoverable
60–75%
Of revenue eaten by founder admin

Five more from our audit library

Anonymised under NDA. Ranked by how closely each maps to ES Group’s reality — field technicians, ERP-to-ERP double-keying, and high-volume admin scaling with job count rather than job value. The #1 below is the audit Olly referenced on the call.

#1 · A 38-person construction/manufacturing business
~1,350 hrs/yrrecoverable · $67K waste tracked
What the audit found: capable, expensive software that didn’t talk to itself. Projects and orders were re-keyed by hand across three separate systems (CRM, project management, accounting). Manual project creation alone was eating 12 hours a week. Ledger reconciliation ran in Excel before WIP could even start. Invoices got double-paid because nothing cross-checked. Drawing up revisions took another 6 hours a week. Across the whole audit, we surfaced upwards of 50 hours a week in recoverable time.
AI opportunities mapped
AI-driven project creation flow across the three systems; automated GL-to-project-ledger reconciliation; AI-native AP automation with cross-checking against the vendor ledger; AI-accelerated proposal/revision generation.
How it maps to ES Group: this is the audit Olly cited on the call — the closest shape we’ve seen to your double-keying problem. They were re-keying across three systems; you’re re-keying across yours plus ~50 different client ERPs. Same fix applies: AI reads what was filed once and writes it everywhere else needed, in each system’s expected format.
#2 · A B2B field-team operation
~7,100 hrs/yrrecoverable · $104K waste
What the audit found: field staff refused to log activity — “they do not want to do it” — so zero visibility on what was happening day-to-day. The 4-system cross-reference (calendar + GPS + accounting + CRM) was confirmed “not working”. The owner ended up spending an entire weekend manually qualifying 150 leads himself because the team hadn’t.
AI opportunities mapped
AI auto-qualification of leads; field-staff activity dashboard with alerts; consolidation of the disconnected systems via AI rather than custom integration; sales playbook built from call patterns.
How it maps to ES Group: the field-staff parallel. Your technicians are in the field doing the work and the admin team picks up after them. Where systems don’t connect, AI bridges the gap by reading and writing on a human’s behalf — no custom integration per client required.
#3 · A multi-team services business
~2,200 hrs/yrrecoverable · $120K waste
What the audit found: the owner personally built tomorrow’s schedule every day, 7 days a week — 3 to 5 hours daily, ~28 hrs/wk on rostering alone. 30 minutes a day reformatting and sending text messages individually to ~50 staff. Lead response sat at 1.5 days, losing about one lead a month at near-100% conversion. “It takes me between three and five hours a day to get it done.”
AI opportunities mapped
Rules-based AI scheduling engine with voice input; CRM lead pipeline closing the 1.5-day response gap; central contact database replacing scattered systems; AI-generated staff communications.
How it maps to ES Group: the coordination layer. Across 25-30K jobs/yr you’re coordinating field technicians, client systems and admin handovers at high volume — the AI scheduling pattern lifts that work to wherever you need it. Bonus from this audit: the recommended new stack was actually cheaper than what they were already paying once an abandoned subscription was cancelled.
#4 · A high-volume customer-onboarding business
~3,500 hrs/yrrecoverable · $253K waste
What the audit found: 300–400 new customers per year onboarded on a fully manual stack — people dragged through pipelines by hand and the same data re-entered across five tools. 15 hours a week across two people went to spreadsheet management alone. Revenue was leaking because only ~1 in 3 customers got properly invoiced. A single automation recovered $47,923/yr of missed revenue — built in half a week.
AI opportunities mapped
Automated invoice & instalment tracking; onboarding automation returning 22 hrs/wk; BI dashboard; central data layer replacing the 5-tool re-entry.
How it maps to ES Group: high-volume operations with manual data entry across multiple systems. Different industry, same pattern. The quick-win lesson is real: one well-chosen AI automation can pay for the whole audit on its own before anything else is built.
#5 · A multi-office sales operation
15,200 hrs/yrrecoverable · $958K waste
What the audit found: 382 steps per transaction across the operation — the owner’s own process map identified 19 hours of AI-replaceable work per transaction. At 800 transactions a year that’s 15,200 hours. No call tracking. No outcome tracking. 200+ Zapier automations held together with no central task board.
AI opportunities mapped
AI to execute the automatable portion of the 382-step workflow; automated outcome tracking; auto task assignment by package type; one-click content generation from templates and data.
How it maps to ES Group: the “same number of touchpoints per job regardless of job value” problem you described. Long workflows with lots of small repetitive steps are exactly where AI compounds — you save time on every single one of your 25-30K jobs.
See a finished audit

Walk through a real audit portal

A complete sample — process map, findings, priority matrix and the full plan. Different industry, simulated data, but exactly the deliverable ES Group would receive.

Explore the live demo →

Piece by piece. Never one big bill.

You stay in control of spend at every step.

The fair questions a co-owner should ask

If you’re forwarding this on, these are the answers he’ll want before Friday.

What about the NZ tax / non-deductibility risk?
Directly addressed above. The audit comes first (no build commitment), the audit itself has a money-back guarantee, and any builds afterwards are scoped one at a time at small price points. No $100K commitments that could become a tax problem if they don’t land.
How does it handle change?
When a client needs new info (e.g. now wants photos too), small changes are configuration on built systems — add a field, drop a field. Bigger changes get scoped. The system’s designed for the real-world fact that clients change requirements.
Are we locked into anything afterwards?
No. You can take the blueprint to any other developer/agency or do nothing at all. The audit is yours either way.
Where does the software live?
Cloud-hosted (Railway or similar) — not on a machine in your office. You own it 100%. Runs 24/7 without you managing infrastructure.
Will our team adopt this?
It’s designed as a co-worker — AI does 95% of the work, your admin team reviews and approves before anything goes to a client. There’s no new SaaS for your team to “learn” in the old sense.
What does our data security look like?
Your data stays inside infrastructure you own. We don’t pool or share data across clients. Detail covered on Friday — happy to send the technical breakdown ahead of the call if useful.

Next step — Friday’s follow-up

On Friday I’ll walk you (and your brother, if he can join) through the short audit plan preview built specifically for ES Group — what the audit would target first, who we’d need to speak to, what it would prove, what the investment looks like, and the specifics of the guarantee.

If anything in here surfaces a question before then, just reply to my email — I’ll address it on the call or send something through in advance.

📅 Friday · 1pm UK time · calendar invite already sent