- Venture buildingPartnering with founders to take mission-led ventures from concept to spinout, at the studio that has spun out nine so far.
- Agentic AIBuilding an agentic delegation product for parents of children with SEND: deciding what agents can safely take off a parent's plate, and what has to stay with them.
- MethodTwo-week sprints against the venture hypothesis, riskiest assumption first, prototypes in front of families early, and evals before anything acts on a family's behalf.
- AlongsideI keep building: prototypes for lawyers, researchers and nurses, and open-source tools for agent work.
Hands-on AI product leader
I ship AI for people who can’t afford it to be wrong.
At Zero Gravity I led the team that took an AI tutor into 850+ UK schools, marking real A-level past papers at over 99% on the syllabus it covers. The government picked it for its AI Tutoring Tools Pioneers Programme, alongside Pearson and ElevenLabs.
- Own
- Build
- Check
- Remember
I am drawn to dauntingly large missions: make death less bureaucratic for families, feed NHS staff through a pandemic, give a student the tutor their family cannot buy. I find the practical problem inside that scale, then turn it into something a team can ship.
Career
- AI productLed six product engineers and a designer on an agentic Socratic tutor on Claude, each engineer owning a problem from prototype to production, and wrote 28% of the commits myself. It went from a 30-school beta into 850+ UK schools, with 47% of monthly users back every week.
- EvalsEvery release passed a golden set of past papers and official mark schemes: bare model 67%, retrieval 79%, our covered content 99%+. The gap was coverage, so we fixed the content. The DfE picked the tutor for its AI Tutoring Tools Pioneers Programme.
- TracesRed-teamed the core failure, a student talking the tutor into the answer, and open-coded hundreds of production traces into failure modes (sycophancy on "I don't know", inconsistency across modes, hallucinated marks). Both became a pass/fail LLM judge on every session, and the Socratic pass rate doubled in two weeks, from 33% to 65%.
- DistributionTeacher-led schools hit 98.7% activation, more than six times the self-signup rate. Pre-built classes and a weekly teacher call became our rollout pattern.
Elliot is the ultimate startup operator — a genuine Swiss Army knife who has made an impact across customer success, operations and product at Zero Gravity.
He doesn't just understand what AI makes possible; he builds with it, experiments rapidly and ships. He now contributes as many PRs as some of our senior engineers, while bringing the customer insight and commercial judgement you would expect from a strong product leader.
If I were starting another business tomorrow, Elliot would be one of the first people I'd try to hire.
Joe Seddon · Founder and CEO, Zero Gravity - PartnershipsScoped, sold and delivered paid programmes inside the early-careers hiring of HSBC, Morgan Stanley, KPMG, Ashurst, Accenture and Snap, owned the milestones and renewals, and doubled the size of those accounts.
- ComplianceISO 27001 and 9001 in under six months as sole DPO, then through due diligence with HSBC, Morgan Stanley, KPMG and Mercedes-AMG F1, with sales cycles 30% shorter.
- OperationsGrew the company from 4 people to around 30, and ran the revenue ops behind £1.5m ARR.
- SupportAn AI assistant resolved 91% of conversations, so a dedicated support hire was no longer needed. CSAT 84%, NPS 80+.
- 2021Probate Provider of the YearBritish Wills and Probate Awards, under 75 staff
- 2021Will Writing Firm of the YearBritish Wills and Probate Awards, national
- Product and opsLed the probate service between customers, legal specialists and the case team, inside SRA and FCA regulation.
- AutomationCase-agent errors down 69% and case handling from two weeks to four days, through HMCTS, HMRC and bank integrations.
- CapacityFixed the input errors eating 30% of solicitor time, so two solicitors did the work of five.
- CustomerNPS from 20 to 80+, and the vulnerable-customer framework the whole company adopted.
MealsForTheNHS
- Zero to oneDay 1 was a WhatsApp group; day 10 was a four-sided marketplace in production.
- Scale303,000 meals in 100 days to 146 hospitals from 223 restaurants, 97% on time.
- GovernanceSet up NHS data-sharing agreements across 146 hospitals, under crisis conditions.
- Team and money£1.8m raised, and a volunteer team grown from 6 to 120 with a hiring playbook other leads could run.
Flash Pack
- 2020Feefo Trusted Service AwardEarned on customer review scores
- 2018Best Young CompanyAmazon Growth Business Awards
- OperationsBuilt CX and crisis operations from nothing: no CRM, ticketing or supplier contracts.
- CustomerNPS held above 90 through 400% year-on-year growth.
- LaunchesThe US launch hit its 12-week plan on the day, and the international partner network tripled.
- Team and safetyHired and trained the first 15 ops staff, with a 100% safety record through the Sri Lanka attacks and COVID closures.
Mentoring
I mentor alongside the products: students and early-career professionals at Zero Gravity, and product peers through Lenny's community.
- I help another operator work through a hard product or career decision, then bring my own.
- I coach students and early-career professionals from low-opportunity backgrounds through university and career decisions.
My stack
The small set of tools I reach for repeatedly.
Pick one to see the job it does in the system; none earns a place here just for being fashionable.
My stack
The small set of tools I reach for repeatedly. Pick one to see the job it does in the system; none earns a place here just for being fashionable.
Writing
When building gets cheap, the handoff becomes the expensive part. On the job that appears when one person carries a problem from prototype to production, and why a rubric has to be uncovered from real failures instead of written up front.
The agent loop is the easy part; knowing when to stop it is the work. The thinking underneath the agent rules on this site's /evals page.
Deploying to first-generation university applicants, people lost in visa paperwork, job seekers in a brutal market. What changes when a hallucination closes a door permanently, read against the EU AI Code of Practice and a run of real failures.
