If you ask an AI model to write your positioning, you get a paragraph that could describe any product in your category. Faster, easier, all-in-one, trusted by teams. That is not positioning. That is the average of the internet. The mistake is asking AI to generate the answer. The right move is to make AI the research and synthesis engine, and keep the judgment for yourself.
Positioning still follows the April Dunford method: understand your competitive alternatives, isolate your unique attributes, map those to the value they create, identify the customers who care most, and choose the market frame that makes your value obvious. AI does not replace any of those five steps. It accelerates the grunt work inside each one. Here is the workflow.
Step 1: Feed it real evidence, not a blank prompt
Generic input produces generic output. The single biggest lever is what you put in. Before you open a chat window, gather raw, specific material:
- Your own win-loss interview notes and sales call transcripts.
- Twenty to fifty of your own product reviews from G2, Capterra, or app stores.
- The same volume of competitor reviews, especially the one-star and five-star ones.
- Support tickets and churn-reason notes if you have them.
Paste that corpus in and ask the model to cluster it, not to conclude. The prompt is closer to: here are 40 customer quotes, group them into recurring themes, and for each theme give me the three most representative verbatim quotes. You are using AI as a pattern finder over evidence, which is exactly what it is good at, rather than as an oracle, which it is not.
Step 2: Mine competitive alternatives the honest way
Dunford’s first step is the hardest to do without bias, because founders underestimate their real competition. Often the alternative is not another tool. It is a spreadsheet, a manual process, or doing nothing. AI helps you surface this if you ask the right question.
Feed it your churned-deal notes and lost-deal reasons, then ask: based on these quotes, what did prospects actually compare us against, including non-software alternatives? The answers frequently include we just kept using Excel or we built something internal. That is the true competitive set, and it changes your positioning more than any feature comparison.
Step 3: Turn attributes into value, and pressure-test it
List your product’s unique attributes, the things competitive alternatives cannot do or do badly. Then, for each attribute, ask the model to argue why a specific customer segment would or would not care. The trick is to ask for the skeptical case. A useful prompt: for each of these attributes, give me the strongest reason a busy operations manager would say so what.
This is where AI earns its place. It is a tireless devil’s advocate. It will not get defensive about your favorite feature. When it cannot construct a compelling value case for an attribute you were proud of, that is a signal your positioning was leaning on the wrong thing.
Step 4: Draft segments, then narrow them
Ask the model to profile which customer types most value the attributes that survived step three. You will get a broad list. Your job is to cut it. The best-fit customer is not everyone who could use the product. It is the segment for whom your unique value is urgent, obvious, and worth paying for. Use AI to generate the candidate segments and to draft a sharp profile for each, then apply human judgment to pick one or two to lead with.
Step 5: Test market frames against a cold reader
Positioning picks a market category that makes your value obvious. Try three different frames and ask the model to role-play a first-time buyer in each. Prompt it: you have never heard of this product. Here is the headline framing it as an X. In one sentence, what do you assume it does and who it is for? Run the same test for each frame. The frame that produces the most accurate, most appealing guess from a cold reader is usually the right one.
The rule that keeps output from sounding like AI
Everything above works because of one discipline: AI synthesizes evidence you supply, and you make every decision. The moment you let it invent claims, invent customers, or write the final positioning statement from nothing, the output collapses to the category average. Bland AI copy is almost always a symptom of blank-page prompting. Feed it your customers’ actual words and it reflects your customers’ actual world back at you.
| Task | Let AI do it | Keep for yourself |
|---|---|---|
| Clustering 50 reviews into themes | Yes | Deciding which theme matters |
| Surfacing hidden competitive alternatives | Yes | Confirming against real deals |
| Arguing the skeptical case for an attribute | Yes | Choosing what to lead with |
| Writing the final positioning statement | No | All of it |
A note for India-built SaaS selling globally
If you build in India and sell to the US and Europe, AI research closes a real gap: distance from the buyer. You cannot casually grab coffee with a mid-market buyer in Austin. But you can feed the model a corpus of US customer reviews and support threads and ask it to explain how that buyer describes the problem in their own vocabulary. It is not a substitute for talking to customers, but it is a fast way to calibrate language and catch the moments where India-market phrasing would land flat abroad.
The workflow in one line
Gather real customer evidence, use AI to cluster and pressure-test it across the five positioning steps, and reserve every actual decision for a human who has to live with the consequences. That is how you get research at machine speed without copy that reads like every other tool in your category.
FAQ
Can AI write my positioning statement for me?
No, and you should not want it to. Asked to generate positioning from a blank prompt, AI returns the average of everything written about your category, which is generic by definition. Use it instead to cluster customer evidence, surface hidden competitive alternatives, and argue the skeptical case for your claims. The final positioning decision has to be made by a human who understands the trade-offs.
What makes AI research output sound generic?
Generic output almost always traces back to generic input. If you give the model no real customer data and ask it to conclude, it fills the gap with category clichés. Feed it actual reviews, call transcripts, and win-loss notes, and ask it to synthesize patterns from that specific corpus, and the output reflects your real market rather than the internet’s average.
Which positioning framework works best with an AI workflow?
April Dunford’s method from Obviously Awesome maps cleanly onto an AI-assisted process because each of its five steps, competitive alternatives, unique attributes, value, best-fit customers, and market category, is a distinct research task you can accelerate with synthesis while keeping the judgment human. Jobs to be done pairs well too, especially for clustering interview transcripts into the outcomes customers are hiring your product to achieve.
