CPG brand · anonymised
Supplier invoices read by AI, written into the ERP.
Order entry, per day
3 hrs → <1
- Time to live
- 4 weeks from kickoff
- Volume
- About 200 invoices a month
Most AI spend never reaches EBITDA. We only build what does.
SECTIONS01 / 10
WE BUILD IT AND WE RUN IT
The gap
Nearly everyone feels AI working. Almost nobody can show it.
That distance — between the feeling and the number — is where budgets quietly disappear. Closing it is the whole job.
Of executives say they're benefiting from AIReported97%
See significant organisational ROIEvidenced29%
68 points of spend nobody can evidence
Source: State of Enterprise AI, Global AI Forum, 2026.0 — 100% of surveyed executives
Clients
Trusted by operators, from growth brands to the Fortune 500.A selection of past and present clients.
Tevine Teas
Fortune 500Several engagements, under NDA
The problem and the fix
Why AI spend fails, and what we do about it.
Every month this runs, the licences renew and the baseline you never took gets harder to reconstruct. We don’t resell platforms or take vendor commissions — if nothing here pays back yet, that is what the value map will say.
The standard
Governance isn’t the brake. It’s what lets you go faster.
Most firms sell governance as compliance — a cost you accept to stay out of trouble. That’s backwards. A board that can approve a use case in a week is the difference between three builds a year and none, and the register is what tells you which ones paid.
- A body that can say yes, no, or not yet — and logs the reason.
- A policy your staff, your customers and your auditors can all read.
- A register that tracks the risk and the payback side by side.
- EU AI Act obligations mapped to what is actually in force today.
Proof
What we can actually show you.
Operating facts, not a maturity-model diagram. The AI practice is led personally — you are buying a named person’s time, not a methodology.
2–3
Builds we recommend, out of dozens mapped
1 day
To a reply from a principal
Zero
Vendor commissions or resale margin
Current
The CTO seat is held today, not in 2019
Engagements
How to work with us.
Questions
Before you get in touch.
How do you measure whether AI actually improved EBITDA?
By baselining before we build. Every use case in the value map carries a current cost — hours, error rate, cycle time, headcount avoided — and a payback estimate against it. After the build we measure the same line. If a use case can't be baselined, that's usually a sign it shouldn't be first.
Why do most AI pilots fail to deliver ROI?
Three reasons, in our experience: the pilot was approved on plausibility rather than payback, no single body owned the decision to scale it, and adoption was treated as training instead of a change to how the work gets done. None of those are technology problems, which is why buying better technology doesn't fix them.
What is an AI governance board, and do we need one?
A standing body that decides which AI use cases are approved, which are off-limits, and how each decision gets logged. You need one once more than one team is experimenting, or once a customer, auditor or regulator might ask what you approved and why. It speeds approvals up rather than slowing them down.
Does the EU AI Act apply to us if we're a US company?
It can — the obligations follow the market, not your head office. But the timeline changed: the Digital Omnibus (Regulation (EU) 2026/1744) entered force on 27 July 2026 and deferred the high-risk regime to December 2027 for standalone systems and August 2028 for AI embedded in regulated products. Article 50 transparency and labelling duties, GPAI provider obligations and the prohibited-practices rules are all live now. We map which of those you actually touch rather than guessing.
How is this different from an AI consultancy?
The person advising you currently holds a CTO seat, so the accountability is present tense rather than a story about a role someone left. We also take no vendor commissions and resell no platforms, which means the value map can honestly conclude that nothing is worth building yet.
We've already tried AI and it didn't stick. Why would this be different?
Because we'd start by finding out why it didn't. Usually the use case was chosen for visibility rather than payback, or nobody changed the process the tool was supposed to replace. We only recommend two or three builds out of dozens mapped, and we stay until your team uses them daily.
Do we need clean data or an in-house tech team first?
No. We design around the systems and data you already have, and the value map accounts for the state they're in — a use case that needs six months of data cleanup gets ranked accordingly rather than quietly failing later. Where a data foundation genuinely is the first job, we'll say so.
How much of our team's time will this take?
For a value map, a few hours per function — we come to your people rather than asking them to prepare for us. The fluency sessions are the exception and are meant to be: leadership has to have hands on the tools, or they can't tell a real opportunity from a vendor pitch afterwards.
0909 / 10Start here
The AI Value Map.
Three weeks, fixed fee. Every opportunity ranked by payback, with a baseline against each — and an honest list of what not to build. Yours whether or not we build any of it.
Request a value mapReply within one business day, from a principal
Contact us
Tell us what you run.
One call with a principal, and a straight answer about where AI would pay in your operation — including when the answer is “not yet”.
- hello@digiform.io
- Response
- Within one business day, from a principal.
- Practice
- AI Transformation
Tell us what you run
Two minutes. It reaches a principal directly — not a queue.
Or email hello@digiform.io directly.