AI advisory

Work out where AI actually fits.

Which tasks it helps with, which tools your people may use, and which data never goes in. Sometimes the honest answer is: don’t build anything.

AI usage policy · what may go whereIllustration
Public materialPublished reports, public web pagesApproved tools, training switched off
Internal documentsDrafts, procedures, meeting notesOrganisation account only
Personal or partner dataStaff, beneficiaries, partners’ filesNever, unless authorised in writing
Illustration. Generic rules showing the shape of a policy. Yours is written for your organisation.

Recent advisory work

Where I’ve done this.

Through Relief Applications, for international and humanitarian organisations. Clients are not named.

AI usage policy and staff awareness

For a humanitarian accreditation body: an eight-section policy, three kinds of account, a green, amber and red table, and eight leadership decisions each with a suggested answer. Plus a staff awareness programme and a one-page guide.

AI usage policyData protectionStaff awareness

AI readiness assessments

For institutions in conservation, development and disaster-risk financing: audit, opportunity map, vendor comparison, recommendations. Supported, remotely, an in-person awareness programme in London.

ReadinessVendor comparison
What an AI usage policy covers: the eight sections +

Front matter

Why now, who and what it covers, what counts as an approved tool.

1 · Principles

Human judgement, no decisions left to AI, confidentiality.

2 · Tools and accounts

Personal, own-business and organisation accounts; where data is processed.

3 · What may be entered

A green, amber and red table, with anonymisation.

4 · Controlled use

When sensitive material may be used: written authorisation.

5 · Use in core work

What’s permitted, recording and transcription, checking results.

6 · Good practice

One conversation per task, memory off, noting AI use, deleting at the end.

7 · Mistakes

What counts as an incident, who to tell, no-blame reporting.

8 · Responsibilities

Who owns what, awareness before use, an annual review.

Closing

The minimum next steps for a pilot, and what it builds on.

The one-week diagnosis

Five days, four deliverables.

€1,500 to €3,000, fixed. No obligation to continue.

  1. Current stateHow the work flows, which tools are in use, where the data sits.
  2. Opportunity mapWhere AI would help, weighed on effort and risk.
  3. Vendor comparisonThe realistic options, including where each processes data.
  4. RecommendationsWhat to do first, what to wait on, what not to do at all.

Where it leads

What the diagnosis can lead to.

“What’s our AI plan?”

Readiness assessment

Where AI would help and where it wouldn’t.

Staff on personal accounts

AI usage policy

What may go into which tool, decided by the kind of data, not the brand.

Sensitive files

AI on your documents

Options compared on permissions and where data is processed.

Rules nobody reads

Staff awareness

A one-page green, amber and red rule your staff will remember.

How it runs

Short, fixed, and written down.

First step, free

A 20-minute call, in English or French.

Diagnosis from €1,500

One week, fixed price. Policy work quoted after a call.

You sign off

I draft; your leadership and DPO decide. Your DPA is signed before any personal data reaches me.

Security & deployment

Start with one question.

What your staff may do with AI, or whether a tool is worth it. Twenty minutes usually tells whether you need a week of work or a page of rules.