Prioritise Organisational AI Investment
Choose between competing AI initiatives with explicit assumptions, dependencies, and running costs: a comparison memo, assumptions register, and staged approvals.
Direct answer
Direct answer: AI investment prioritisation
When a leadership team has more AI proposals than money or attention, the useful question is not which idea sounds most transformative but which one has the clearest owner, the most testable assumptions, and the dependencies the organisation can actually meet. I help leadership teams compare initiatives on the same terms, record what each assumes, and approve them in stages so the next tranche of funding depends on evidence rather than enthusiasm.
The question
“How should leadership choose between AI initiatives when benefits, dependencies and operating obligations are uncertain?”
Who it is for
CEOs, leadership teams, and boards allocating budget across several AI proposals, pilots, or vendor offers.
What you leave with
An initiative comparison memo, an assumptions register, and staged approval decisions.
- 01 A side-by-side comparison of each initiative: outcome, owner, data and system dependencies, operating obligations, and reversibility
- 02 An assumptions register listing what must be true for each initiative to pay off, and how it will be tested
- 03 A staged approval plan: what is funded now, what evidence unlocks the next stage, and what stops it
- 04 A short list of initiatives deliberately deferred, with the reason recorded
How it runs
Format, method, and preparation
- Duration
- Preparation interviews with initiative sponsors, a half-day prioritisation workshop, and a written memo
- Delivery
- In person in the UK or remote
- Participants
- The leadership team or board, plus the sponsor of each initiative under consideration
Put every proposal in the same shape
Restate each initiative as an outcome, an owner, the change in work it requires, and what it depends on, so they can be compared fairly.
Surface the assumptions
Separate what is known from what is hoped: data quality, adoption, vendor capability, integration effort, and the cost of running the system after launch.
Count the operating obligations
Make visible the ongoing work each system creates: monitoring, evaluation, review queues, incident handling, and retraining people.
Decide in stages
Agree what to fund now, the evidence each initiative must produce before its next stage, and the conditions under which it stops.
Useful to have ready
- →The proposals, business cases, or vendor quotes under consideration
- →Access to each initiative sponsor before the workshop
- →A rough indication of the budget and capacity available
Evidence
What this draws on
Six companies founded, 2008–2024
Kwippy, Jaja.tv, Octo.ai, ExpressMOJO, Boom, and Neul Labs, spanning social software, interactive media, ML and analytics, logistics, blockchain infrastructure, and applied AI. Advisory and investment records are not counted as founded companies.
Read more →Machine learning at Hike
Built or led NLP, recommendation, computer-vision, and trust-and-safety ML systems serving more than 100M users. Company-level outcomes are not attributed to one person.
Regulated banking AI at Aveni
Architected enterprise banking AI with conduct-risk workflows, evidence generation, human review, escalation, evaluation, versioning, and release controls. Part of the Aveni team in the first FCA Supercharged Sandbox cohort; the FCA lists Aveni as an accepted firm, which is not an FCA endorsement or certification.
Inspect the source ↗Scope and limits
What this is not
- —The memo supports the leadership team’s own decision; it is not a financial forecast, valuation, or investment advice.
- —Benefit estimates remain the sponsors’ estimates. The work makes their assumptions explicit and testable; it does not validate projected returns.
- —I do not recommend vendors I have a commercial relationship with without disclosing it first.
If the need is different
Common questions
Answers before you commission
Our leadership team disagrees on which AI initiatives to fund. Can a workshop resolve that?+
It can turn disagreement into explicit criteria, priorities, and owners. Disagreement usually comes from unstated assumptions about value, effort, or risk; writing them down side by side makes the trade-off discussable.
How do you compare AI initiatives with very different benefits?+
By putting each into the same structure: outcome, owner, dependencies, operating obligations, reversibility, and the assumptions that must hold. Comparing what each needs to be true is more reliable than comparing headline benefit figures.
Should we build, buy, or partner?+
That choice is made per initiative, not as a policy. The comparison records what each option means for data access, control, switching cost, and ongoing operation, so the decision can be revisited when circumstances change.
What does staged approval mean in practice?+
A first tranche funds the cheapest work that tests the riskiest assumption. Further funding is released only when agreed evidence exists, and the stop conditions are written down before the work starts.
Related