TL;DR
- The setup fee is not a service. It is a hedge against your data being worse than the vendor hopes, packaged as a product because the vendor is not allowed to sell consulting.
- Traditional industry buyers do not evaluate AI the way the technology press does. They ask how much grunt work disappears, whether the quote goes out faster, and whether NPS moves.
- Dogmatic packaging loses for both sides. The buyer cannot trust a templated onboarding with a career riding on it, and the vendor leaves real value unpriced.
- Comparative quotes suggest one-time builds with incremental consulting cycles are cheaper for the enterprise than a license from a company running on a grow-or-die clock.
I promised to share the lessons as I learn. This is one such, from my recent, intense field conversations.
Three days on the dance floor
A few weeks ago, I met 300+ traditional, old-school industry executives. They were all from wildly successful companies ($100M - $2B) that still fondly remember Lotus Notes and ask you to do a Skype call. I loved how they pretend-danced on the floor with 25+ AI product companies that tried to woo them with AI ERPs, AI CPQ software, and what not.
There were even a couple of robotic-sounding IVR demos which we had to pretend-acknowledge that they were sounding human.
I loved that dance floor. I was in it for three days. It was better than the ‘LaCroix’-powered Databricks conference in the valley where everyone wore a t-shirt size smaller than they should. It is not an important detail that I like the boomer industry song and dance party because they were cutting checks.
Or were they?
Where the pitch goes wrong
I spent some time with a handful of YC companies (the gold-standard startup pixie-dust that gets you $10M seed round) at the event. Between crushing some croissants, I saw them pitch their wares. Boomer CXOs were intently taking notes on their cliffpads.
What does a typical license cost?
(Uhm..no one was ready with an answer and they were air dropping one quickly to fill the silence.)
Well, we need to study your catalogs, digitize them, harmonize the data and then configure our agentic workflow. So, I guess a set up cost of $100,000.
Can I ask you how many SKUs you have in your catalog?
That must be a couple of millions.
Oh jeez, that’s a lot. I think you are looking at $200K neighborhood for annual cost apart from the one-time set up.
This is when I smell blood. This is dogma selling dogsh*t. I know exactly why startups do this. The products are merely demos. They are not enterprise-ready. The $100K setup fee is a ‘Hail Mary!’ to ensure they just don’t go under the water trying to untangle the data mess.
What they really want to say is ,“Bob, we want to do a workshop to understand your goals, review your data, present a data maturity assessment, and then we will go from there.”
God, I missed Accenture in the room. They would have sold a $2M deal right there.
Instead of pitching a consulting service, the startup packages the goo into a set up fee. More goo gets packaged as annual contract because who knows how many workflows, what tokens, and how the company wants the toast buttered.
It’s either this or a set up fee and a $0.5 per item in the catalog or some such ‘outcome or consumption’ sounding pricing.
What the buyer is actually assessing
Sadly, enterprises don’t study tech from YC or Twitter/X. Some of these industries don’t even know Gartner from Gutenberg. So their assessment I simple – how much grunt work am I saving; does it improve my NPS; am I generating a quote faster, and suggesting substitutes early on – all without doing an internal ping-pong between supply, pricing, and sales teams.
They are skeptical about this whole thing working but they smell cost-savings. The VC-orchestrated, dogmatic school of SaaS–pricing–trained founder has some unseen boundaries they play within. They will only sell if it’s a product and professional services is a no-no. Outcome and consumption pricing are new-fangled, imaginative half-way midpoint where they meet the industry. Well, the industry does not care.
What should have happened is:
- A two-day workshop, followed by
- Data and Process/workflow assessment (with as much digging as possible, into what $ figure outcome that this initiative can generate)
- A cost + margin internal quote based on FTE, AI costs and a value-based quote that ‘the client won’t fall off the chairs hearing it’
- Followed by a 'negotiation song-and-dance’ not just for the pricing but how the program can be stacked in cycles of incremental de-risking through intermittent, verifiable outcomes (and not some meta, end-of-the-chain outcomes)
The price could be more or less or equal to the packaged ‘Setup and recurring annual contract’ pricing from the SaaS company, but pricing isn’t the point. It is the approach, setting up for success, and capturing value.
Dogmatic approaches are lose/lose for both the sides:
- A risk-averse enterprise buyer may not trust a startup and it only gets worse when transparent-sounding-but-useless pricing communication and templated onboarding is how far a product company is willing to go. A failed program is a failed career for the buyer. Worst yet, it’s how AI gets shut off from the org.
- The startup or the product company leaves a lot of value on the table because they won’t do what is right but they would have to fit everything into a neat setup and onboarding cycle
Well, I am glad we took a consulting approach to Moative because we anticipated exactly this. Enterprises need someone at their conference room in Oklahoma presenting and reviewing phased implementations over onion burgers and pecan pie – every two weeks. They need to feel in control. They need to see a steering committee presenting ‘red, yellow, and green’ traffic signals.
While I said that the pricing does not matter as much as the approach, I have a good feeling (informed by comparative quotes) that it’s a lot more affordable for enterprises to do one-time builds and incremental updates through consulting cycles delivered through specialist AI shops than buying from a star-spangled YC company that runs with a ‘grow or die’ motto.
The enterprise on the other side worked on a ‘Never Die’ motto since World War I and their fasteners are still the same and their catalogs are still in paper files.
Happy Independence Day, America!
Where Moative fits
Moative works the way these buyers need to buy: assessment first, phased delivery, verifiable outcomes at each step, and a steering committee that can see the status without a translator. In healthcare revenue cycle management that means proving the denial rate moves on one payer before anyone signs up for the enterprise rollout. The related argument about who should end up owning the resulting system is in Buying vs. Becoming AI.
Frequently asked questions
Why do AI startups charge a large setup fee?
Because the work required to make the product function on your data is consulting, and the venture playbook discourages selling consulting. The fee is that work, repackaged, plus a margin for the risk that your data is worse than expected.
What do traditional industry buyers actually evaluate?
Grunt work removed, cycle time on quotes, NPS, and whether the process stops bouncing between supply, pricing and sales. Not model benchmarks, and not architecture.
Is consumption or outcome pricing better than a license?
It is an improvement in intent and usually still the wrong shape. Both assume the value is in the product rather than in the assessment, redesign and integration that make the product work in your environment.
Why does templated onboarding fail in the midmarket?
Because a failed program is a career risk for the buyer, and a template offers no way to de-risk it in stages. Without visible intermediate outcomes, the buyer has to bet everything on one delivery.
Is buying AI software cheaper than a build?
Often not. A one-time build with incremental consulting cycles frequently costs less over a few years than a license that never falls in price, especially while the underlying model costs keep dropping.
How should an AI program be structured for a cautious buyer?
In cycles, each ending in something verifiable that the buyer can show upward. Intermediate outcomes, not a single outcome at the end of the chain, and a cadence regular enough that nobody has to wonder what is happening.
