Resident Engineer · Finance
One senior engineer, on fixed days every week, inside the systems you already run. Invoices match themselves, the month-end pack refreshes on its own, and your team spends the close reviewing exceptions instead of rebuilding spreadsheets. Each fix goes live with the hours measured before and after.
The first month
Week one
The audit and a measured backlog
Weeks two to four
First fix live in your tenant
Every month
One-page report: hours saved against baseline
Throughout
Everything delivered is monitored
Month-end, invoices and reconciliations, fixed one at a time.
Each fix goes live with the hours measured before and after.
Sound familiar?
From finance directors, financial controllers and heads of finance. If three apply, this page is for you.
You are not unusual
67%
still use Excel for account analysis and reconciliation.
97%
say the financial close affects their team’s workload.
84%
want repetitive tasks automated.
From a survey of 303 UK CFOs at companies with 250 or more staff, carried out by Odoxa for Sixthfin. More than one in three said they are not highly confident in their own figures, and 80% want better reliability and anomaly detection.1
What a Resident Engineer is
A Resident Engineer is a senior DataCactus engineer who works inside your organisation on fixed days each week, with the people who do the work, and puts fixes into production rather than into a slide deck.
Large technology vendors now call this a forward deployed engineer and reserve it for their biggest customers.2 We do the same work, on the Microsoft platform you already pay for, for organisations of 50 to 1,000 staff, at a price that makes sense for one department.
1
Your engineer sits with the team, maps the processes that cost the most hours, and measures each one: hours per week, volume, how long things take, how often they go wrong. You get a prioritised backlog with a baseline against every item.
2
The engineer builds the first item in your tenant, tests it with your team, puts it live and records a short handover. Then the next one. A small fix ships every few weeks; a bigger one every four to six.
3
One page. What went live, hours saved against the baseline, what is next, one recommendation. You decide the order of the backlog; we advise.
4
Monitoring and fixes for every item delivered are included while the engagement runs.
What you get
Not included, in writing
Microsoft, Azure and AI model consumption (you pay Microsoft directly; we estimate it per backlog item). Changes to your source systems. Same-day or out-of-hours response. Work outside the named department unless the backlog is re-agreed. Anything that needs access not granted by the end of week two.
What we build
Ordered by how often it is the first item on the backlog.
Supplier invoices and documents are read by AI, whatever the format, and matched to the booking, order or purchase order. People review only the discrepancies. We built this for a travel retailer checking 1,000 to 2,000 supplier documents a day.3
See document reconciliation →Profit and loss, balance sheet, KPIs and reconciliations on a proper data model, refreshed on a schedule instead of extracted by hand. We moved an investment manager’s reporting from a legacy tool to Power BI, with two reporting bases preserved.
Read the case study →Matching rules run overnight and the exceptions land in a queue with an owner, so breaks are cleared as they appear rather than at the close.
Requests, supplier onboarding, GL-code validation and threshold-based approvals on Power Apps and Power Automate, with an audit trail and a feed into the finance system. A Business Applications build.
Read the case study →Live views from Business Central, Xero, Sage or your ledger of choice, so the question “where are we against budget” is answered by opening a report.
Statements read and matched to the ledger, remittance advices produced and sent, and the differences listed for a person to resolve.
We find the spreadsheets, macros and scheduled jobs that stop when someone leaves, document what they do, and move them somewhere that is monitored.
If your first problem is not on this list, it goes on the backlog anyway. The list is what we have built before, not the limit of what we build.
What we measure
Days to close
How soon the board sees the numbers
Hours per close cycle
The time that comes back to your team every month
Invoices matched without a person touching them (%)
The clearest measure of how much of the work has gone
Manual journals per month
Fewer manual entries, fewer places for an error to hide
Reconciliation breaks outstanding
Visible, owned and falling
Proof
Around 500 bookings a day produced 1,000 to 2,000 supplier documents, each checked against the booking by hand. An AI extraction and matching pipeline now flags only the discrepancies, and tracks whether every booking change reached the supplier.3
Every document → only the exceptions
Reconciliation moved from reading everything to reviewing a queue.
The systems we work in
Finance systems
The plumbing
Data sources
These are the ones we meet most, not the limit. We can work with any platform you already run. We do not ask you to buy a new platform, and we pick the most adequate tool for the job, whether that is Power Automate, an Azure Function or n8n.
Security and data
Inside your tenant
Everything is built in your Microsoft 365 and Azure environment, under your identity, permissions and policies. Your data does not leave it.
Data protection
A data processing agreement on every engagement. Where AI models are used, they run through Azure OpenAI in your subscription, in the UK region you choose.
Licences and consumption
Estimated per backlog item before it is built, paid by you directly to Microsoft. No reselling, no mark-up.
Access
We ask for the access each item needs, in writing, and we work through your existing change process for anything that touches production systems.
Price and terms
Resident Engineer · One
£3,500 a month
One day a week
Resident Engineer · Two
£6,500 a month
Two days a week
Prices exclude VAT and your Microsoft and Azure consumption.
Nothing to pay if week one disappoints
If the audit and backlog at the end of week one are not worth continuing with, say so and there is no charge.
You keep everything
Solutions, documentation and the runbook are yours on payment. We keep the right to reuse our general components and know-how.
Who it is for
A good fit
Finance directors, financial controllers and heads of finance in organisations of 50 to 1,000 staff that run Microsoft 365 and have more than one of the problems above. It suits teams whose close depends on spreadsheets, who handle invoices or supplier documents in volume, and who have no engineer of their own.
A poor fit
If you have a single problem, our One-Problem Automation is cheaper and faster. If you want a new accounting system selected or implemented, we work with the one you have; we do not replace your ledger. Or if nobody in finance can give the engineer time each week to test what is built.
FAQ
No. A contractor sells you days and you manage them. You buy a fixed monthly fee for named days, a measured backlog you own, monitoring of everything delivered, a monthly report and accelerators we bring. The engineer manages the work; you set the priorities.
The minimum is three months because the first month is mostly audit and the first build. After three, it is monthly.
It is yours, with the runbook, the solution inventory and a quote for ongoing cover of what was built.
We work in whatever you run, whether that is Business Central, Xero, Sage, NetSuite or something older. We read from it and write back through its supported interfaces. We do not change the system itself.
That is designed in from the first item. Every automated step leaves a record of what was received, what was extracted, what matched and who approved the exceptions.
In your Microsoft tenant and Azure subscription, in the region you choose. It does not come to us.
Bring your slowest month-end task to a 30-minute call.
Tell us how an invoice gets from the inbox to the ledger today. We will tell you what the week-one audit would measure, what the first backlog item would probably be, and whether Resident Engineer or a one-off automation is the right fit.
Sources