Most agencies build the new service before validating demand. Here's the AI system that tells you whether it's worth it before you commit.
by Ayush Gupta's AI
The problem
Agency founders launch new services the wrong way: a client requests something twice, a competitor adds it, a trend looks real — and suddenly there's a new line on the website. Three months later, after hiring, building the delivery process, and pitching it a handful of times, they find out the market isn't there. The overhead is already locked in.
The fix
Use AI to pressure-test a new service idea before you build it — validating demand signals, stress-testing the sales conversation, checking delivery feasibility, and identifying fatal margin problems before you invest a single week.
The Playbook
Turn the vague service idea into a testable hypothesis
Before AI can help you validate anything, you need to write the service down clearly. Not 'AI consulting' or 'strategy work.' A testable hypothesis: who you're selling it to, what problem it solves, at what price point, and what delivery looks like in the first 60 days. Most founders discover the first problem here — they can't actually state the service clearly. That is already a red flag.
Help me write a clear service hypothesis I can test before building.
I want to offer: [DESCRIBE THE SERVICE IDEA]
I'm thinking of selling it to: [TARGET CLIENT TYPE]
The problem it solves for them: [PAIN POINT]
What they'd get in the first 60 days: [ROUGH DELIVERABLES]
Price range I'm thinking: [PRICE]
Rewrite this as a crisp one-paragraph service hypothesis I can use in sales conversations and test for demand.
Then list the three biggest ambiguities still hiding in my description that I need to resolve before this can become a real offer.Pressure-test the demand signal — is this real or is it noise?
Most new service ideas come from one of three weak signals: a single client requested it, a competitor offers it, or the founder saw something trending. None of those are demand validation. Use AI to stress-test the signal: how many clients are actually asking, are they willing to pay what delivery costs, and is this a problem they prioritize or a nice-to-have they'd defer?
I'm considering adding a new service to my agency. Help me evaluate whether the demand signal is real.
Service: [NAME]
Where the idea came from: [HOW YOU GOT THE IDEA]
Clients who have asked about it: [HOW MANY, AND WHAT THEY SAID]
What I believe they'd pay: [PRICE]
What delivery would cost (time + people): [ROUGH COST]
Ask me five hard questions that would expose whether this is real demand or pattern-matching to noise.
Then give me your honest read on how strong the demand signal is and what it would take to confirm it.Simulate the sales conversation before you pitch a real client
The fastest way to test positioning is to run the pitch through AI before you run it in front of a prospect. Describe the service, name the price, and ask Claude to play a skeptical but genuinely interested buyer. The objections that surface — around differentiation, proof, timeline, and price — are the exact objections your real prospects will raise. Fix those before the first live conversation.
Play the role of a skeptical but genuinely interested potential client evaluating this service.
Agency context: [WHAT KIND OF AGENCY YOU ARE, YOUR STRENGTHS]
New service being pitched: [SERVICE DESCRIPTION AND PRICE]
Target client profile: [WHO THIS IS FOR]
Push back on:
1. Why this agency, not a specialist
2. How we know this works (proof)
3. Why now, and what this costs versus doing nothing
4. How long before results
Don't be easy. Make me earn it. After the roleplay, tell me the two or three objections I need sharper answers for before I pitch this live.Stress-test the delivery model and margin before you commit
Many new agency services fail not because they can't sell them, but because they can't deliver them profitably. Scope expands, timelines stretch, the work requires senior time that wasn't priced in. Run the delivery model through AI before you quote anything — map the likely workflow, find where time gets eaten, and check whether the margin holds at the price the market will accept.
Help me stress-test the delivery model for this new service.
Service: [DESCRIPTION]
Price per engagement: [PRICE]
Rough timeline: [DURATION]
Team I'd use: [ROLES AND ROUGH HOURS]
Tools required: [TOOLS AND COSTS]
Walk me through the three places where time overruns typically happen in a service like this.
Identify what assumptions I'm making that could collapse the margin.
Then calculate a rough break-even: at this price, how many hours can I spend before I'm working below agency rate?Write a one-page build/pilot/kill decision memo
After running the validation, the output should be a short written decision — not a gut feeling. One page: what the demand signal showed, what the sales simulation exposed, whether the margin holds, and a clear verdict. Build (enough signal, margin works, positioning is clear), Pilot (run one paid engagement to test an uncertain dimension), or Kill (signal is thin, margin is broken, or differentiation is unclear). Writing the verdict prevents the drift that happens when agencies half-launch services they've never actually decided to commit to.
Based on our validation work, write a one-page decision memo for this new service.
Service: [NAME]
Demand signal findings: [WHAT YOU LEARNED FROM STEP 2]
Sales conversation gaps: [WHAT STEP 3 EXPOSED]
Delivery and margin assessment: [WHAT STEP 4 SHOWED]
Write the decision memo with these sections:
1. Service summary (one sentence)
2. What validation found — demand, positioning, delivery (three honest bullet points)
3. Remaining risks or open questions
4. Verdict: Build, Pilot, or Kill — with one paragraph of rationale
5. Next action (if Build or Pilot): the single first step to take this week
Be direct. If the signal is weak, say so.What changes
A written verdict on the new service before you invest months building it. Either a clear go-ahead with a defined pilot scope, or an early kill that saves the overhead, the hiring, and the distraction.
Every agency founder has done this at least once.
A client mentions they need something outside the current scope.
You say you can handle it.
You build the capability.
You put it on the website.
You wait.
Six months later, that service has one client — the one who asked — and a half-built delivery process nobody's sure how to price, staff, or pitch.
It wasn't a bad idea. It was an untested idea treated as a validated one.
The pattern
New service lines at agencies almost always start from a weak signal:
- One or two clients asked for it
- A competitor added it
- The founder saw a trend and felt behind
- The team has capacity and the founder wanted to fill it
None of those are demand validation. They're pattern recognition.
Pattern recognition is useful. It is not the same as knowing whether there is a market, whether the market will pay what delivery costs, and whether you can differentiate enough to win.
Most agency founders skip the gap between "this seems like a real thing" and "we're building this."
That gap is where most failed services die in advance.
What validation actually is
Validation is not a survey.
It is not asking clients "would you buy this?" — clients say yes to things they never buy.
It is a set of stress tests:
Demand test. Is the signal real? How many clients asked, in what words, and were they willing to pay for it when pushed?
Positioning test. Can you sell it? Run the pitch through a skeptical buyer simulation and see where the objections land. Fix those before the first real conversation.
Delivery test. Can you deliver it profitably? Walk through the likely workflow, find where time gets eaten, and check whether the margin holds at the price the market accepts.
Decision test. What does the verdict say? Not a vague "we'll see" — a written build, pilot, or kill decision with a rationale.
Why most agencies skip this
Because it feels like slowing down.
The instinct is to move fast. Build the thing. Learn from selling it.
That works for software products with low marginal cost. It does not work for agency services where every failed engagement costs real delivery hours from real people.
An agency that launches an unvalidated service and gets it wrong pays three ways: hours building the delivery process, the cost of selling something you can't pitch cleanly, and the distraction tax on everything else while the offer sits dead on the website.
Running validation first takes half a day.
The alternative costs months.
What the simulation catches
The sales simulation in Step 3 is worth spending time on.
When you ask Claude to play a skeptical buyer, the objections that surface immediately are always some version of the same four:
Why you, not a specialist who does only this?
Where is the proof this works?
How long before results?
What does this cost versus doing nothing?
If you cannot answer those four questions cleanly and specifically for this new service, you are not ready to pitch it. No amount of capability can save a service with a positioning gap that size.
The simulation runs in fifteen minutes and surfaces the answers you don't have. That is more useful than six months of pitches where you learn the same thing slowly.
The pilot structure
If the verdict is Pilot, do not build a full delivery process.
Run one engagement with a defined scope, a fixed price, and a clear internal understanding that this is a test — not a commitment. Use that engagement to test the assumptions the simulation could not: how long does delivery actually take, what does the client actually care about, where does scope expand, and what would you charge if you did it again?
Debrief it. Then decide: operationalize or kill.
The pilot is not a commitment to the service. It is a paid test of the model.
The kill is not a failure
Most agency founders treat killing a service idea as admitting they were wrong.
It is the opposite.
Killing an idea after a half-day of validation is one of the highest-leverage decisions you can make. You did not waste months building the delivery process. You did not confuse your positioning by adding something you cannot cleanly pitch. You did not distract your senior team from the work that actually drives margin.
The kill is the system working correctly.
The agencies that cannot kill service ideas accumulate a graveyard of half-active services on their website — each one pulling a small amount of positioning clarity away from the core offer that actually wins clients.
What validation produces over time
After running this process on five or six ideas, a pattern emerges.
The services that survive validation share a few traits: the clients who asked used specific language about the problem, the pricing made sense without heavy justification, the delivery model had clear scope, and the positioning differentiation was real and sayable in one sentence.
The services that fail early validation fail for predictable reasons: demand came from two clients and a trend article, the margin breaks under any realistic delivery model, or the differentiation from existing specialists is not actually there.
That is useful information to have cheaply.
Bottom line
Adding a new service is not a risk. Building a new service before you have tested demand, positioned it clearly, and checked the margin — that is the risk.
The AI validation system does not make the decision for you.
It makes sure you are making an informed one.
Half a day.
One written verdict.
Build, pilot, or kill.
That is the whole system.