Early-Stage AI Product Investment Scorecard
Rank AI startups by defensibility, use-case value, market potential, and team quality.
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 6
- Confidence
- 94%
The scorecard evaluates an early-stage AI company across several connected dimensions: the importance of its use case, its ability to find product-market fit, its path to scale, its capacity to evolve beyond the initial feature, and its defensibility against competitors or platform vendors. Reviewers first assess the customer value demonstrated by the product, then examine whether the market is broad enough to support a substantial business. They explicitly model threats such as OpenAI or incumbent analytics products absorbing the feature. Founder quality is treated as a necessary diligence input rather than an afterthought. Finally, opportunities are ranked relative to one another, and capital is concentrated according to conviction instead of being distributed evenly merely to avoid making a hard choice.
Origin
Extracted from Marketing Against The Grain during an AI Shark Tank review of five early-stage AI products.
Core principles
- 01Judge the underlying opportunity, not just the quality of the demo.
- 02Prefer important use cases with credible paths to product-market fit.
- 03Test whether the company can scale and evolve beyond its initial feature.
- 04Assess defensibility against platforms and established competitors.
- 05Investigate the founders before committing capital.
- 06Concentrate capital when the evidence does not support equal allocations.
How to run it
- 1
Define the investment test
Write down the criteria that will determine whether a company deserves investment before comparing the candidates. Include defensibility, use-case importance, product-market-fit potential, scalability, and room to evolve.
Pro tip Frame the test around which company could become a breakout business within a specific period.
Watch out Changing criteria after seeing a favorite demo encourages confirmation bias.
- 2
Evaluate customer value
Determine whether the product solves an important problem and whether the demonstrated use case creates meaningful value. Separate an impressive interaction from a business-critical outcome.
Pro tip Ask whether customers would pay for the outcome rather than merely enjoy trying the feature.
Watch out A magical demo can conceal a weak or difficult-to-monetize use case.
- 3
Test market breadth and evolution
Assess the total addressable market and identify adjacent use cases that could expand the business. Look for a credible path beyond the initial beachhead.
Pro tip Favor products whose core capability can support multiple valuable workflows.
Watch out Do not assume a broad technical capability automatically produces a broad commercial market.
- 4
Model defensibility threats
List platforms, incumbents, and adjacent products that could reproduce the feature. Decide whether integrations, workflow depth, data, distribution, or brand could protect the company.
Pro tip Distinguish short-term distribution opportunities from durable competitive advantages.
Watch out A head start is not defensibility if a platform can make the feature native.
- 5
Investigate the founders
Evaluate the founders' domain knowledge, speed, judgment, and ability to attract customers and talent. Treat missing founder information as an unresolved diligence gap.
Pro tip Ask what the team understands about the market that competitors do not.
Watch out Do not make a final investment decision from product information alone.
- 6
Rank and allocate by conviction
Compare candidates directly, select the strongest opportunities, and allocate capital according to the evidence. Challenge any allocation that spreads money evenly without a company-specific rationale.
Pro tip Require a written explanation for both the selected company and the strongest rejected alternative.
Watch out Diversifying equally across every candidate can disguise an unwillingness to make a decision.
In the wild
The hosts favored Kajoo because converting sketches and Figma designs into code addressed a broad, transformational category. They also liked Deepsheet's natural-language analytics, but identified direct threats from OpenAI Code Interpreter and analytics vendors adding similar interfaces. Their final ranking therefore placed Kajoo ahead of Deepsheet despite strong enthusiasm for both products.
→ Kajoo received the hosts' hypothetical angel check, while Deepsheet remained the runner-up.
ChatGPT initially split a hypothetical $100,000 evenly across all five companies. After being challenged to justify that weak allocation, it concentrated $60,000 in Kajoo and $40,000 in Deepsheet, citing greater defensibility, broader use cases, and larger potential markets. It also identified founder information as the key missing diligence input.
→ The revised allocation reflected differentiated conviction and exposed an important information gap.
Common mistakes
Confusing demo quality with business quality
A polished demonstration can make a narrow or weakly monetizable use case appear more investable than it is. Evaluate the durable customer outcome separately from the presentation.
Ignoring platform substitution
A useful feature may be absorbed by OpenAI, an analytics vendor, or another incumbent. Explicitly assess whether the startup can retain an advantage after competitors copy the interface.
Allocating evenly without conviction
Equal allocations can look prudent while avoiding the central task of ranking opportunities. Weight investments according to evidence and explain every allocation.
Is it for you?
Best for
It is best for angels, seed investors, and operators comparing several young AI products with limited operating data.
Not ideal for
It is not ideal for pricing mature companies that already have extensive financial and market-performance records.
From the transcript
“I think it's defensible, I think it's a really good idea and an important use case, I see an opportunity to actually find product market,…”
“It said they were more defensible, the use case is broader, probably a bigger TAM.”
“The one thing I would really want to know that I don't know is I would want to know more about the founders.”
From the episode
AI Shark Tank: 5 Founders Pitch Us Their Ai Products (#124)