Hello, I would be grateful for a sanity check on two eligibility points before I submit, because both are touched by refusals reported here this year.
Profile
- About 15 years or more in software engineering and enterprise data systems before 2023.
- I made a deliberate move into applied AI in 2023 after being promoted to manager: I applied through my employer’s internal job market to transfer into a specialist AI/data science group, then took educational leave for a self-funded MSc in a related STEM field (math and data science) at a UK university. I returned at the end of 2024 and have worked hands-on in applied AI since early 2025. I have subsequently earned 12 vendor certifications in AI, data, and cloud through rigorous continuous study, including intermediate-level ones.
- Employed by a large global professional-services firm. Although my grade is at manager level, my work and focus remain strictly hands-on: I write code, and I design and evaluate systems myself. I do not do sales or advisory work.
Key implementations since 2025
- A natural-language data agent for the field sales organisation of a global life-sciences company. In that product team I was the only engineer on the AI/agent workstream: agent architecture, semantic-layer design, test strategy and test-case design (covering every branch of the agent’s logic), automated/custom evaluation with ground-truth preparation, and load testing. Released to production in mid-2026 to over 500 field users; 95% of survey respondents use it weekly.
- An LLM-based job-application essay proofreading and feedback feature in a university’s student app. I designed the feature and implemented its prompts, while separately designing the evaluation method and building the framework to run it (the review points, the ground-truth set and the metric choice). It went live in September 2025, and the product team kept using my evaluation set-up after I left.
- A complete modernisation of a legacy ML forecasting tool. I redesigned its core architecture into a scalable ELT pipeline on a managed data platform. By directly addressing user pain points, I automated tedious manual code edits and local runs, drastically reducing processing time for the business. I managed this as a user-centric product enhancement, rigorously verifying accuracy parity before production deployment.
- UI/UX design and front-end build of the human-review screens of a generative-AI drug-safety tool.
Question 1: “early stage in their career”
Background:
- The guide says Promise applicants are “likely to have less than 5 years of experience in technology, but can have had a longer career in another type of work”.
- I have read the advice given here this year that applicants with 8-15 years in tech should not apply under Promise. My earlier years were in software engineering, so I understand that “another type of work” may not cover me in the way it covers someone coming from, say, finance or healthcare.
- My reading is that production LLM-agent engineering is itself a field that has only existed for a few years, and that my career in it began with a proactive and documented change of direction in 2023, with hands-on work from 2025. I will show the earlier years openly, as a separate “earlier career” block.
Questions:
-
Q1.1: Is that defensible in practice, or will assessors treat any prior software engineering as “experience in technology” and count it in full?
-
Q1.2: If the latter:
-
(a) Is Exceptional Talent a realistic alternative for someone with no external recognition (no talks, publications, awards or open source)?
-
(b) Are there other ways of framing or evidencing an early-stage profile that have worked for people in a similar position?
Question 2: consulting work, and OC3
Background:
- I have seen a refusal letter quoted here (August 2026) containing the sentence “consulting and agency work do not qualify for any of the criteria”. On the other hand, I am also aware of several successful cases where applicants employed by consultancies have been endorsed. Everything I have built was for clients of my employer. I understand that the employer itself is not product-led, and that “technical applicants from non-technical organisations are eligible”.
Questions:
-
Q2.1: What I cannot tell is how far that sentence reaches for OC3 in particular (and for the other criteria, if you know). Does it mean:
-
(a) advisory/consulting work does not count, but an engineer who built and shipped a product that real users use can still present that product as evidence; or
-
(b) anything built for a client by a consultancy is excluded, no matter how technical, or no matter how large its commercial contribution or impact?
-
If (a), what helps an assessor see the difference: system logs and code history showing my own work, user numbers, public documentation of the product, or short confirmations (in email form) from third parties such as colleagues or client staff?
-
If (b), I would rather know now.
-
Q2.2: OC3 is worded “founder, senior executive, board member or employee of a product-led digital technology company”. I have seen recent posts and refusal extracts saying that software delivered to clients on a bespoke basis does not count. For a technical employee of a professional-services firm whose products are used by the firm’s clients:
-
(a) Has OC3 ever been accepted in that situation, and on what kind of evidence?
-
(b) If not, is OC3 effectively closed for such applicants, so that they have to rely on OC1 plus another criterion (OC2 or OC4)?
What I have already read
The application guide in full, the GOV.UK pages, the rejection-pattern threads, and the consultancy-related and Promise-eligibility threads from 2025-26. I am not asking “will I pass”. I am trying to find out whether either point should stop me, or whether there is still a realistic chance and it is worth trying.
Thank you for your time and any advice you can share.