Step 1: Decide what you are actually selling
Most applied AI companies are not selling a model; they are selling a better way to finish a job — reviewing contracts, answering customer tickets, forecasting stock. Write down the job, the buyer who owns the budget for it and what they use today. If the honest answer is 'a general assistant for everyone', narrow it before you raise.
Investors increasingly assume foundation models will keep getting cheaper and better. Your pitch has to explain why that helps you rather than replaces you.
Step 2: Lock down your data position
Before you open a round, audit where your training, fine-tuning and evaluation data comes from. For each source, record the licence or contract, whether customer data can be used to improve the product, and how personal data is handled under UK GDPR.
- A data inventory listing every source and the rights attached to it.
- Customer contract wording on data use and model improvement.
- A short note on how proprietary data compounds as you grow.
Step 3: Show defensibility beyond the model
Wrapping a third-party model is a legitimate start, but you need a story for what becomes hard to copy: workflow integration, proprietary feedback data, distribution, regulatory approvals or deep domain knowledge. Show at least one of these already forming, not just planned.
Step 4: Build unit economics that include inference
Model a single customer: what they pay, how many requests they generate, what each request costs in model and infrastructure fees, and what support they need. Show gross margin today and your plan to improve it — caching, smaller models, routing or your own fine-tuned models.
Step 5: Choose the right funding route
Pre-revenue teams usually raise from angels, accelerators and pre-seed funds, often using SEIS or EIS in the UK so investors get tax relief. Once revenue is repeating, seed and Series A funds become realistic. Some founders use revenue-based finance or grants for specific R&D. Pick the route that matches your evidence, not the one with the largest cheque.
Step 6: Run a tight investor process
Build a list of investors who have backed AI at your stage, prepare a short deck and a data room, and run conversations in a concentrated window of a few weeks so momentum builds. Track every conversation, follow up quickly and share updates while the round is open.
Fundraising readiness checklist for AI startups
| Area | Ready when you have | Common gap |
|---|---|---|
| Problem and buyer | A named budget owner and a paying or committed customer | Broad 'AI for everyone' positioning |
| Data rights | Documented licences and customer consent to model improvement | Scraped or unclear data sources |
| Defensibility | Evidence of a moat forming now | Relying on prompt design alone |
| Unit economics | Per-customer margin including inference | Model costs left out of the plan |
| Data room | Deck, financial model, cap table, contracts, data inventory | Assembled after investors ask |
Questions to prepare before you pitch
- What happens to your product if the underlying model provider cuts prices by 90% or launches a competing feature?
- Which data do you have that a new entrant could not easily get?
- What is your gross margin per customer after model costs?
- How do you measure output quality, and how do you catch failures?
- Why is now the right time for this product?
How KJ Enterprises evaluates applied ai businesses
KJ Enterprises backs applied AI businesses that solve a specific, paid-for problem and have a credible plan to keep margins healthy as they scale. Through AutoThink Group we build and operate AI products ourselves, so we look closely at data rights, quality control and how the product fits into the customer's day-to-day work.
If that describes your company, you can apply for investment. Related reading: the complete applied AI funding guide and how to raise for a machine learning company.
Frequently asked questions
Do I need revenue to raise money for an AI startup?
Not always. Pre-seed investors back strong teams with early proof such as pilots, letters of intent or usage data, but they expect a clear plan to reach paying customers.
Is building on a third-party model a problem for investors?
Not by itself. Investors want to see what you add on top — data, workflow and distribution — and how you would cope if the model provider changed pricing or terms.