One That Got Away: Data & AI Foundations for a Venture Portfolio
Before FIELD STATION existed, we put this proposal to an impact accelerator. We didn't get the work, but it shows the shape of what we want to do together, so we're publishing it with the client anonymised.
We wrote this proposal last month, before FIELD STATION was incorporated. The client was an impact-focused venture accelerator that supports a portfolio of social-purpose ventures and operates in partnership with a university. They put out an invitation to quote for a combined data and AI review; we responded jointly. They liked what we suggested, but we didn't get the work.
We're publishing it anyway, with the client anonymised, for two reasons. First, it's a good example of the kind of engagement we want to keep doing: data, knowledge and AI treated as one connected piece of organisational development rather than three separate audits, with the people who'll use the findings involved in producing them rather than waiting for a report to land. Second, we said when we started FIELD STATION that we'd build in the open, including the parts that don't land the way we hoped. A proposal that didn't win the work is one of those parts.
The text below is close to what we submitted. We've replaced the client's name and any identifying detail about them with generic descriptions; everything about our own approach, team and past work is unchanged.
The brief
An impact accelerator and a university partner invited quotations for a six-month Data & AI Review, with a budget of up to £30,000. The work was split into two connected tracks: a data mapping exercise to inventory venture and ecosystem data, identify gaps and governance risks, and produce a phased implementation roadmap; and an AI review to assess current adoption, maturity, capability gaps and responsible-use guardrails, then prioritise high-value use cases.
A third thread ran through both: integration. The brief asked specifically how information flows across platforms, what operating inside the university's technology environment means for tooling and controllership, and whether the current architecture can carry AI-enabled ways of working. The successful partner would work closely with a senior sponsor and a Data & AI Working Group.
Our reading of the brief
This review builds the data and AI foundations for the accelerator to track what its ventures are achieving, learn from its portfolio, and make better decisions about where to direct support. We treated it as one connected piece of organisational development, rather than two separate audits.
How we met the brief
| Brief requirement | Our response |
|---|---|
| Data landscape & mapping | Mapping of the venture portfolio led by use cases, alongside ecosystem and network, operational, and university-connected data flows. |
| Knowledge management | Analysis and documentation of how insight, venture learning and decision-relevant knowledge are captured, stored, shared and reused. |
| AI adoption review | An audit of current tools, informal ("shadow") use, capability, risks, opportunities and readiness for more deliberate adoption. |
| Governance & compliance | Practical recommendations on stewardship, ownership, GDPR flags, responsible-AI guardrails and the relationship with the university partner. |
| Strategic roadmap | A prioritised 12–24 month roadmap across data, knowledge and AI, with sequencing, owners and dependencies. |
| Working Group capability | Collaborative sessions, shared methods and handover materials so the group can continue the work. |
Our approach
- Questioning. We start from the decisions the accelerator needs to answer, then follow them into the data, knowledge and AI that support or obstruct them. Instead of simply producing an inventory, we map what matters.
- Integrated. Data/knowledge and AI run in parallel and inform each other. Adoption depends on data quality and confidence, and data strategy now has to account for AI use.
- Capability-building. The Data & AI Working Group should shape findings, test analysis and practise the methods throughout, so capability and ownership stay with the client.
- Value-based. We provide neither hype nor alarm: analysis of where AI can help, where it should be constrained, and the conditions needed for responsible use.
- Knowledge in the open. Insight and venture learning are assets in their own right, so we look at how they are captured, stored, shared and reused, not just where data sits.
Methodology
The Data & AI Working Group is a participant, not an audience.
We proposed a six-month engagement across five phases, aligned with the kick-off timeline set out in the brief. The data/knowledge and AI tracks run in parallel and come together through shared sense-making, prioritisation and roadmap development.
| Phase | Focus | What happens |
|---|---|---|
| 00 | Initiation & alignment (weeks 1–2) | Agree priority questions and the in/out line; gather existing material; take an early read on the technology and governance relationship with the university partner; set up the Working Group as a participant. |
| 01 | Parallel discovery (months 1–2) | Use-case-led mapping of venture, ecosystem, operational, knowledge and university-connected data, alongside an AI usage assessment (tools, informal use, capability, risks) via interviews, light surveys, workshops and walkthroughs framed to surface real practice safely. |
| – | Go / no-go checkpoint | Confirm the highest-value areas before committing the second half. |
| 02 | Analysis & sense-making (months 3–4) | Data flows, gaps, quality and GDPR flags, plus a pragmatic, conceptual core entity model; AI maturity, a shortlist of high-leverage, lower-risk use cases and draft responsible-use guardrails. Validated with the Working Group. |
| 03 | Synthesis, roadmaps & governance (month 5) | One prioritisation framework (value, effort, risk, readiness); a 12–24 month roadmap of quick wins, foundational and medium-term actions with owners and dependencies; proportionate governance. |
| 04 | Handover & capability (month 6) | Final synthesis, a senior decision workshop, Working Group handover and leave-behind templates with clear answers on what to do first, what to avoid, and who owns next. |
We worked through three questions the brief raised:
- How information flows across platforms.
- What operating inside the university partner's technology environment means for tooling, controllership and procurement.
- Whether the current architecture can carry AI-enabled ways of working.
What we would produce
Rather than separate documents for each indicative output, we proposed to consolidate these into artefacts that would help the client make decisions:
- Data, knowledge and integration mapping: a data inventory and asset register, a venture lifecycle data map and an ecosystem data map, ownership, quality and gap analysis, GDPR flags, an integration assessment with the university partner, and a core data model with entity definitions.
- AI assessment and responsible-adoption framework: an AI usage and maturity assessment (including informal use), capability findings, an AI needs and opportunity map, a prioritised AI use case portfolio, and lightweight guardrails.
- Integrated roadmap and governance model: one prioritisation framework, governance and ownership recommendations across data and AI, and a 12 to 24 month roadmap with owners, sequencing and the Working Group's future role.
- Integrated findings report, decision workshop and Working Group pack: a senior synthesis report, a facilitated decision session, and a resource pack (glossary, governance templates, facilitation guide).
Ways of working
- Cadence. Fortnightly check-ins with a senior sponsor inside the accelerator, monthly Working Group sessions, and a shared live workspace for documents, actions and emerging findings.
- Open by default. We share drafts early and think out loud rather than disappearing and returning with a finished report. This means the client sees the work as it forms.
- Remote-first, but in person when it counts. Most work runs remotely; we prioritise in-person time for kick-off and the final decision workshop.
- Confidentiality and data handling. Venture data is often commercially sensitive, so we treat everything we see in discovery as confidential, hold it only for the life of the engagement, and return or delete it on completion.
Team
Tom Watson Data strategy, governance and organisational data management. Advisor, technologist and builder across data infrastructure, open standards, impact and insight tooling, AI adoption and organisational resilience. Leads delivery, the data/knowledge landscape, the university integration, the core model and the roadmap; co-leads synthesis and facilitation.
Dr Doug Belshaw AI literacies, capability and facilitation. Consultant and facilitator helping social-purpose organisations build AI literacy and responsible practice, through TechFreedom and direct client work; earlier roles at Mozilla, BBC R&D, MIT and City & Guilds. Leads the AI literacy and capability strand and co-leads facilitation and synthesis.
We build, not just advise. Real, in-use data and AI tools: Bearing (open AI-model recommendation), llmstxt-social and the RACE Report data platform (The Good Ship), and a whole suite of AI-created and informed tools, including Commonplace, Sightlines, and Substrate (Dynamic Skillset). That gives us practitioner judgement about what actually works, scales and lasts.
Relevant experience
Every example below is work with a social-purpose organisation, and much of it sits in the same portfolio and venture-support context this client works in.
Longitude Prize on Dementia (Social Tech Trust): data and AI support to a portfolio of 12 ventures, from early-stage start-ups to established organisations, for a body that sits between funder and accelerator, much like this client. The work was this client's own role in miniature: strengthening data infrastructure and AI practice across a mixed portfolio, and helping the portfolio learn from itself.
Local Needs Databank (NPC), ClientEarth & Joseph Rowntree Foundation: organisation-wide data, impact and insight infrastructure, covering user research, data mapping, shared standards and working tools that turn distributed data into intelligence rather than reporting.
CASORT: supporting the team through the shift from prototypes to a mature product function: the practices, structure and decisions that move from "it works in a demo" to something dependable, close to the in-house capability this client wanted to build.
BBC Responsible Innovation Centre & TechFreedom: Doug co-authored the BBC R&D's AI literacies framework; together we run TechFreedom, helping social-purpose organisations think clearly about technology dependence and responsible AI. Tom sits on the Ada Lovelace Institute Community Forum.
Budget & payment
Fixed fee of £30,000 excluding VAT for approximately 40 days of combined senior time over six months, paid in four equal stages of £7,500: on signature, on completion of discovery, on completion of analysis, and on final handover.
| Area | Indicative allocation |
|---|---|
| Project management, initiation & coordination | £3,500 |
| Data, knowledge & integration mapping | £9,000 |
| AI review & maturity assessment | £7,500 |
| Synthesis, roadmap & governance | £5,500 |
| Working Group enablement & handover | £3,500 |
| Contingency | £1,000 |
| Total | £30,000 |
Assumptions, risks & dependencies
- University technology environment: we assumed autonomy over tools, data controllership, GDPR, procurement and integration could enable or constrain the roadmap.
- Honest AI disclosure: informal use tends to be under-reported if people feel audited, so we designed for "safe surfacing".
- Coordination: if the client was also commissioning ecosystem-insights work elsewhere, we'd want to coordinate to avoid duplication and make the outputs reinforce each other.
- Timely access: we assumed timely access to people, documentation and systems from kick-off, and would agree a realistic Working Group time commitment at initiation.
If this is the kind of review your organisation needs, get in touch. We'd rather do this work than have it sit in a drawer.
Get in touch: contact@fieldstation.xyz