Thinking of Joining an AI Startup? 12 Red Flags and the Questions to Ask First

Joining an AI startup: lessons and red flags from a real story

A recent post on Reddit titled “I wish someone had warned me before I joined this AI startup” has been doing the rounds. The author describes joining an early-stage company full of optimism and leaving a few months later feeling burned out, excluded, and ultimately dismissed with little warning.

It’s one story, one perspective. But it surfaces patterns plenty of people will recognise in the current AI gold rush: unclear products, impossible KPIs, long hours as standard, and a culture that treats people as disposable. For UK readers, there are also legal and ethical considerations around pay, contracts, and data practices that are often overlooked in early-stage chaos.

What happened, in brief

“There was no real onboarding or clarity on what the company was actually building.”

The poster describes being assigned growth targets (thousands of sign-ups) for a product that wasn’t fully defined or ready, with minimal guidance on strategy. Workloads were intense (55-60 hours per week), expectations kept ratcheting up, and internal communication felt fragmented. The author says they were let go abruptly, and alleges previous employees hadn’t been paid – a serious claim if true.

Key details like runway, headcount, and leadership background were not disclosed.

Why this matters for the UK tech scene

AI is attracting funding and talent at speed, but velocity without governance creates real risk. In the UK, basic obligations still apply – paid work must be paid, written terms are required, and data must be handled lawfully. High-growth experimentation is no excuse for ignoring employment law or misrepresenting what’s built.

For candidates and founders, this is a reminder: clarity, structure, and respect matter at Seed and Series A just as much as they do later on. They’re also a competitive advantage in a tight talent market.

12 red flags before you join an AI startup

• No clear product or customer – lots of talk about “AI” with no defined problem, ICP (ideal customer profile), or active users.
• “Growth before fit” – aggressive sign-up targets for a product not yet shipped, not used, or not keeping users.
• No onboarding or documentation – no 30/60/90 plan, no roadmap access, no clear decision-makers.
• KPIs that ignore reality – top-of-funnel goals with no budget, no channels, no assets, and no definition of “qualified”.
• Always-on culture – 55+ hour weeks “as normal”, weekend work implied rather than exceptional.
• Opaque decision-making – private chats, shifting priorities, and big decisions without context or retrospectives.
• Vague compensation and equity – no written offer, unclear option terms, or no mention of EMI in the UK.
• Contractor misuse – pushed into “contractor” status with employee-style control and hours, potentially raising IR35 risks.
• Late or missing pay – any suggestion of “we’ll sort payroll later” or stories of unpaid ex-staff.
• Data compliance as an afterthought – no mention of UK GDPR, data minimisation, or security when handling customer data.
• Founder behaviour – badmouthing ex-employees, pressuring people in public, or framing burnout as “hunger”.
• Runway and governance secrecy – refusal to share basic runway estimates, investor updates, or who sits on the board.

Questions to ask before you accept an offer

Product and customers

• What problem are we solving, for whom, and what alternatives do they use today?
• What’s live right now, how many active users do we have, and what retention looks like?
• How do we define product-market fit and what evidence would we accept?

Role and success measures

• What are my 30/60/90-day goals? What resources, budget, and access will I have?
• How will we prioritise if targets and product reality diverge?
• Who makes decisions and how are trade-offs documented?

People and culture

• How do we run stand-ups, planning, and retros? What’s written vs verbal?
• What was the biggest mistake last quarter and what changed as a result?
• Can I speak to future teammates and at least one former employee?

Compensation, equity, and terms

• What’s the base, bonus, and equity package? Are options under an EMI scheme and what’s the vesting schedule?
• Is the role employed or contractor? If contractor, why, and how do we handle IR35?
• What’s the company’s runway and top three risks? How often do you share updates?

Data, ethics, and compliance

• How do we handle UK GDPR – lawful basis, data minimisation, and security?
• What’s our policy on production data, AI model prompts, and logging?
• Do we have an incident response plan and named data protection lead?

UK-specific due diligence checklist

• Verify the company exists and is filing on time at Companies House.
• Check you’ll receive a written statement of employment particulars within two months: GOV.UK guidance.
• Ensure salary meets National Minimum Wage. Genuine interns are rare; most “intern” roles must be paid.
• Understand option grants and EMI: GOV.UK EMI overview.
• Ask how personal and customer data are handled and review UK GDPR basics via the ICO.
• If pushed into contractor status, read up on IR35 and who bears risk.

If you’re already in a bad situation

• Document everything – offer letters, timesheets, messages, and payslips.
• For unpaid wages, start with ACAS Early Conciliation. You can also use Money Claim Online if needed.
• If you suspect minimum wage breaches or unlawful deduction from wages, see GOV.UK: if your employer doesn’t pay you.

This isn’t legal advice – speak to a qualified adviser if you’re in dispute.

Healthy signals to look for instead

• Clear and modest targets tied to what’s actually shipped and learned.
• A written 30/60/90 plan and agreed metrics with realistic budgets.
• Regular planning and retros, with decisions recorded and shared.
• Transparent runway updates and a straightforward equity package (ideally EMI in the UK).
• Basic hygiene on data protection: access controls, retention, and incident plans.

Practical tip: make targets measurable and visible

If you do join, insist on clarity and instrumentation from day one. Build a simple funnel model, define what counts as a qualified signup, and track weekly. A lightweight workflow using Sheets can go a long way – I’ve written a guide on how to connect ChatGPT with Google Sheets to automate analysis and reporting.

Final thought

Early AI startups can be brilliant places to learn and build. They can also burn people out when urgency substitutes for leadership. The Reddit post is a reminder to ask hard questions, do basic checks, and walk away if answers aren’t forthcoming. There’s plenty of exciting work out there; you don’t need to compromise on fundamentals to do it.