Services
Operations Intelligence is the core of the practice, delivered as Vernier: one checked view across the systems an organisation already runs, and an assistant you can ask about it. It is delivered in four stages, in order: roadmap, pilot, build, retainer. Most start at the roadmap because it is the cheapest way to find out whether the rest is worth doing. All of it is done by the same people, so the plan never promises what the build cannot deliver.
Vernier
A central operations intelligence engine. It connects the data your organisation already has, checks each source against the others, and turns what it finds into actions a person approves. Any industry, any source with an export or an API, in your own Google Cloud or AWS account.
- 01
Connect
Read from the systems you already run, through their exports or APIs. Nothing is replaced.
- 02
Check
Match the same customer, site, product or asset across systems, then compare. Where sources disagree, the gap is recorded, not averaged away.
- 03
Surface
Each morning, every role sees its part: what changed, what is out of line, and what it is costing.
- 04
Act
The follow-up is drafted. A person approves anything that moves money or changes a record.
Vernier is not tied to an industry. It works on whatever the organisation records.
- SalesTills, e-commerce, billing
- FinanceERP, ledger, payments
- CustomersCRM, support, marketing
- PeopleHR, payroll, time
- OperationsTickets, inspections, forms
- Assets and IoTSensors, meters, telematics
- PartnersSupplier and platform exports
- DocumentsSpreadsheets, registers
Vernier is running in production today, built for loss prevention and governance. Every view answers one question: can this number be checked against another system or a physical count? It covers nine areas, all checked against each other.
- Sales and transactionsDiscounts, voids and corrections by operator
- Third-party platformsReported sales checked against payouts
- StockCounts against theoretical stock, valued
- Supply chainFill rate, short deliveries, stock in transit
- LabourSchedules compared with timecards
- PeopleHR and payroll, private reads logged
- Repairs and equipmentTicket log, quotes approved by a second person
- PropertyTurnover rent checked against actual sales
- FinanceProfit per site; supplier bills approved twice
The store app
Scoped per systemCounts, deliveries, waste and the daily checklist on the store's tablet, with recount checks and notices from head office.
- Replaces
- Paper counts and checklists typed up later, with errors found weeks after they mattered.
- Needs from you
- Your store and region structure, your stock list, and the checks you want each store to run.
Case study not published yet.
Aggregator reconciliation
Scoped per systemOrder-level reporting across your third-party aggregators, and a check on whether the payout matched the sales the aggregator claimed.
- Replaces
- Weekly exports that are complete enough to file and too coarse to question.
- Needs from you
- Access to each aggregator's reporting export, and a decision on how far back to load history.
Know the sales numbers
Six to ten weeksYour point of sale connected to an assistant, so the operations team asks in plain English and gets figures traced to the revenue centres underneath.
- Replaces
- Reports nobody has time to run, and questions nobody bothers to ask because the answer takes a day.
- Needs from you
- Whatever your POS will expose. That is the constraint, more than the assistant is.
Repairs and maintenance ticketing
Scoped per systemFault reporting, vendor assignment, cost and store sign-off as an installable web app: one ticket log for every store.
- Replaces
- Faults raised by phone, message and email, with nobody able to see what is open, who has it, or what it cost.
- Needs from you
- Your triage rules, and a decision on who is allowed to assign work to which vendors.
- 01RoadmapFour weeks
Workshops with the people who lead the work and the people who do it. Ends with a ranked roadmap, a list of what to leave alone, and the first pilot scoped and priced.
Fixed price. A period of due diligence, ending with the first build scoped and priced.
- 02PilotScoped in the roadmap
The first thing worth building, built. Small enough to judge on its own, and chosen in week four rather than guessed at in week one.
Priced in the roadmap, before you commit to it.
- 03BuildSix to twelve weeks
The system in production, in your own cloud account, with documentation good enough for someone else to take it over.
Fixed scope, priced against the saving.
- 04RetainerMonthly
Support with a stated response time. Standard on every build, because a system in production needs someone watching it. You can still take the documentation and run it yourself.
Monthly. Cancel when it stops earning its place.
Most start at the roadmap, because it is the cheapest way to find out whether the rest is worth doing. If you already know what you want built, say so on the call and we start at the build.
What sits inside each stage.
Six services across the two kinds of work. The roadmap, governance and enablement are the planning side; the rest is what gets built.
Workshops with leadership and with the people actually doing the work. We list candidate use cases, score each on feasibility, risk and return, and rank them. The first pilot comes back scoped to a fixed price.
You get- Prioritised roadmap
- Pilot scope and price
- Maturity snapshot
- A list of what we would leave alone
Right for you ifYou know AI should be doing something here, but not what, and you do not want to spend to find out.
A risk assessment of what is already happening in your business, a governance framework sized to your organisation rather than a multinational, and the documentation an auditor will ask for. We write policies people will actually read.
You get- Risk register
- Governance framework
- Responsible-use policy
- POPIA, GDPR and ISO audit pack
Right for you ifA board, an auditor or a client has started asking questions you cannot answer yet.
Role-specific sessions for finance, operations, sales or support: what these tools do well, what must never be pasted into one, and how to check an answer before you act on it. Your team leaves with material they can use that afternoon.
You get- Training sessions
- Prompt library for their actual work
- One-page internal usage policy
Right for you ifAdoption is uneven, output quality is a lottery, and nobody has told anyone what the rules are.
Delivered as Vernier. We connect the systems you already run, from sales and finance to CRM, HR, operations and IoT, and check each source against the others. Where they disagree, the gap is recorded rather than averaged away. A morning briefing summarises what changed, an exceptions list puts a cost and an owner against each gap, and an assistant answers questions with every figure linked back to the rows it came from. Access is by role, so leadership sees the whole and a team lead sees their area.
You get- One checked view across your systems
- Morning briefing and exceptions list
- Assistant with source-linked answers
- Role-based access and usage log, in your own cloud
Right for you ifThe answer exists somewhere in your systems, but getting it takes a person and a day, because three of them disagree.
Reporting and reconciliation, stock counts, repair tickets, document digitisation, and the glue between systems that currently lives in somebody's inbox. Built as proper software: logins, audit trails, offline tolerance, no per-seat licence that punishes you for growing.
You get- Working system in production
- Documentation and handover
- Support retainer, standard
Right for you ifA weekly process eats a person's week, and everyone has stopped noticing.
An assistant connected to mail, files, calendar and tasks. It reads and drafts on its own, proposes actions, and executes only what a named person approves. Built on a provider-neutral core, so the model underneath can be swapped without a rebuild.
You get- Deployed assistant in your cloud
- Approval and audit layer
- Web and mobile access
- Model-swap tested
Right for you ifYou want the leverage, but nothing may send, pay or delete without a human in the loop.
Who this works for,
and who it does not.
The industry matters less than the shape of the problem. It works best where the data already exists but sits in systems that do not agree: sales in one, finance in another, customers in a CRM, the rest in spreadsheets. If a routine question takes a person a day, because the figures have to be pulled from three places and reconciled by hand, this pays.
It is a poor fit in three cases, and we would rather say so on the first call than three weeks in. A business that already runs on one system answering its questions has little to join. A customer-facing chatbot on a website is not what we do. And if the honest answer to what happened last month is that nobody wrote it down, there is nothing for a system to stand on; that is a records problem first, and a cheaper one to fix.
What four weeks
actually looks like.
Our discovery engagement, week by week. Nothing here is a workshop for its own sake: by week three there is a rough working version of the hardest idea, so the roadmap is grounded in something that ran.
- Week 1
We listen
Sessions with leadership and with the people doing the work. We watch the process rather than ask about it.
- Week 2
We score
Every candidate use case gets a feasibility, risk and return score. Some get struck off, and we say why.
- Week 3
We test the hard one
The top candidate gets a rough working version against your real data, so the roadmap is not a guess.
- Week 4
You decide
A ranked roadmap, the first build scoped to a fixed price, and a clear list of what we would not touch.
How we price.
No day rates you cannot forecast, and no licence that punishes you for adding people.
Fixed price, agreed before we start.
There is no standard cost, because there is no standard system. A period of due diligence is what establishes it. You commit to that, not to what follows, and it ends with the first build scoped and priced. If the scope changes, we re-quote rather than run a meter.
Fixed project scope, priced against the saving.
Where the outcome is measurable, we price against what it recovers or saves. Where it is not, we say up front what we will measure instead.
Monthly retainer with a stated response time.
Standard on every build, because a system in production needs someone watching it. Cancel when it stops earning its place. The handover documentation is still written so you can run it yourself. That is the point of writing it properly, and nothing is held hostage either way.
Everything.
Systems run in your Google Cloud or AWS account, on your data, with documentation good enough for someone else to take over. Nothing is held hostage, and the AI model underneath can be swapped without a rebuild.
Not sure which track you need?
That is what the call is for. Describe the work; we will tell you whether it is a plan or a build, and what it would cost to find out.