Identifying fit-for-purpose AI tools and models

Purpose: Ask AI which AI tool, model, or research modes are best for the task at hand

• Prompt

Finding an underserved ICP

Purpose: Ask AI to perform desktop research (web search; academic literature review) to identify any niche target audience segments with unmet needs

• Prompt - Underserved niche ICP
• Prompt - ICP that answers “why now” question

Analysis of first responses (Claude 3.7; ChatGPT 4o & 4.5)

Claude 3.7’s final recommendation ‘Port Authority Managers & Harbor Operators’ was based on one single data source (the fact that a tech player in this space had recently received funding). When I asked it to provide more data to build its case for this ICP, it provided a few extra sources with very weak evidence. Its main line of argumentation? “Two other startups received funding in this space, and their founders say there are massive unsolved pain points.” Not strong enough for me.

I had similar experiences with ChatGPT (trying with both 4o and 4.5, web search on).

ChatGPT’s argumentation for the headers Pain Point; Outdates Solutions in use; SaaS willingness came from a 5-month old LinkedIn post by Daniel Kalish which details what he learned after speaking to 20+ CFO’s. This is anecdotal, at best (For clarity: I’m certainly not shtting on Kalish here, only on ChatGPT!).*

I compared the full responses from Perplexity Pro using Claude; ChatGPT 4o Mini, ChatGPT 03-mini and ChatGPT 4.5, and decided to continue my conversation with ChatGPT 4.5, which gave me the most satisfactory first response.

Sparring with ChatGPT 4.5 didn’t go without hiccups. I pointed out that his evidence was weak

• Each argument was backed up by only one source, not multiple
• Sources not existing (make sure to click each link!), biased (e.g. startups in the space pointing out what a great space it is), stale, generic, or otherwise unreliable

“Please re-argument your case for (niche xxx) being the winning niche ICP. Try to name at least 3 sources (hyperlinked) for each argument you make. The evidence should be strong, relevant, and as recent as possible. This means that reports from 2025 weigh more heavily. Fact check yourself before responding.”

Now shift your focus to niche ICPs (Ideal Customer Profiles) that have developed a relatively new and urgent need — in other words, ask yourself: 'Why now?'

For example, consider how many social, health, and arts organizations recently lost funding due to Trump-era budget cuts and the reduction of USAID support. That created a sudden gap, and a timely opportunity for new solutions.