Advice

Clarity before
commitment.

Independent advice on where AI pays off and where it plainly doesn't. No vendor pays us, so the recommendation follows your problem rather than a margin.

Business process

Which processes actually benefit

We map how the work really moves — not how the process diagram says it does — and find where time is lost. Some of it is worth automating. Some is worth restructuring. A good deal is best left alone, and we will say which.

  • Where the queue actually forms, measured rather than assumed
  • What AI genuinely improves, and what a simpler change fixes faster
  • A shortlist ranked by payback, not by how interesting it is
Data management

Getting the data fit to use

Most AI disappointments are data problems wearing a costume. Before anything is built we establish what you hold, who owns it, how sensitive it is, and how long you are allowed to keep it.

  • What exists, where it lives, and which copy is authoritative
  • Classification and retention, so confidential material is handled as such
  • An honest read on whether the data can support what you want from it
Cloud or local

Which route fits which task

Microsoft, Google, or a private deployment on hardware you control. Each wins on something and loses on something else, and the honest answer is usually a mix rather than a single winner.

  • Judged on confidentiality, time to value, running cost and control
  • Per workload, because the answer for a shared inbox is not the answer for a client archive
  • We implement all three, so the recommendation costs us nothing either way
Client feedback

Hearing what clients actually say

Feedback arrives scattered across mail, calls, surveys and complaints, and is usually acted on by whoever shouted loudest. Read at volume it becomes something else: themes, with a trend attached.

  • Every channel read consistently, not just the formal survey
  • Recurring themes separated from one-off irritations
  • A direction of travel, so you can see whether a fix actually worked
Legal aspects

What it exposes you to

The EU AI Act, the GDPR and professional confidentiality all bear on what you may deploy and on what terms. We work out which obligations actually apply to your use, and what has to be true before it goes live.

  • Where your use falls under the AI Act, and what that requires of you
  • Lawful basis, processor agreements and what the vendor may do with your data
  • Professional privilege and confidentiality, treated as a hard constraint
AI governance

Who may use what

External rules tell you what is permitted. Governance decides what happens inside: which tools your people may use, on which material, and who signs off. Without it somebody eventually pastes a client document into a public chatbot.

  • A policy short enough that people actually read it
  • Approval routes for the decisions that warrant one, and none for those that don't
  • A register of what is in use, so shadow tooling surfaces early
Enablement

Making it stick after we leave

The most common failure is not a poor model but a good system nobody uses. Adoption is a design problem: who it helps first, how they learn it, and who owns it once the engagement ends.

  • Training built around the actual work, not a feature tour
  • A first group who benefit visibly, because adoption spreads from them
  • A named owner internally, so it does not decay the moment we leave
Cost and return

What it costs, and when it pays back

Boards ask this first and rarely get a straight answer. We model it properly: licences, inference, hardware and the staff time it actually consumes, against the time it gives back.

  • Total cost at real volumes, not the pilot's bill
  • The break-even point stated plainly, with the assumptions behind it
  • The option of not proceeding, costed alongside the rest
How advice runs

Short, specific, and yours to keep.

01

Look

We spend time with the people doing the work, not only the people who commissioned the review. What we find is what we report.

02

Assess

Each candidate is weighed on payback, risk and effort — including the option of doing nothing, which sometimes wins.

03

Recommend

A short written recommendation with the trade-offs made explicit, in language your board and your engineers can both act on.

04

Hand over

You keep the analysis. If you want us to build what we recommended we will, and if you would rather someone else did, the report is written so they can.

Bring us the decision you are stuck on. You will get a straight answer, including when that answer is that AI is not the right tool for it.

Ask a question