Act 1
[Scene: Staff meeting, seven execs sitting around conference room table, eighth exec on Teams]
CEO: It's time for us to get with AI! Everybody, AI as fast as you can!
Other leaders: Ok! How do we do that? I love AI!
CEO: I don't know, can we give everyone a chatbot?
Leaders: Ok! Chatbots all around! Also, there are a lot of new vendors in my inbox. Give me a huge budget for experimentation and I will try every single one of them!
CFO: Ok! That sounds like a good plan to me. We have been living in a ZIRP world and money is basically free. I bet that will go on forever! Let's get with AI!
[fade to black as person on Teams finally manages to unmute]
Act 2
[CEO calls meeting to order]
CEO: I can't wait to hear how it's going with all those AI tools we bought!
Leaders: Actually, new idea: we've studied these tools for a year and it turns out we can build them all ourselves! After all, coding is just vibes now. So we can definitely replicate the workflows that specialist vendors have spent decades mastering. We'll just need two summer interns.
CFO: Amazing! This will save us so much money. I will even give you a third intern to make this happen!
CEO: I like what I'm hearing!
Leaders: LFG!
[fade to black, person on Teams sends thumbs-up emoji as execs file out of room]
Act 3
CEO: Board meeting next week. Tell me about all the things we've transformed with AI.
Leaders [uncomfortable]: Check out this usage data!
CEO: Great! But where are the business results?
Leaders: Look, the usage curve is up and to the right!
CFO: And so is our AI spend!
CEO: I can't go to the board with this. We promised them AI transformation, not AI spend!
Everyone [in unison, eyes darting around the group]: What are we going to do?
Consultant [off stage left, walking in with superhero cape]: Did someone call my name?
[fade to black, curtain falls]
Build vs. buy vs. old-fashioned advice
Over the last 24 months, AI transformation has exploded. I don't just mean exploded like "has gotten really big." I mean exploded like, for organizations that are getting results, "AI transformation" is completely different work compared to when they started:
Two years ago: Companies were running wild with experimentation budgets. The AI tools they piloted didn't work that well yet, so they mostly did not renew.
One year ago: People decided to build their own solutions. They discovered it's hard to vibe code an entire ERP in-house. (Shocker.)
Six months ago: Operations jobs in specific areas like "AI enablement" and "AI optimization" were replacing jobs in the more generic "AI transformation." Companies were making significant in-house investments in technical leaders (or at least technical-ish leaders) who were tasked with helping a company transform its operations with AI. I took a role like that last year myself.
These days: Disappointed by all of the above approaches, people are increasingly turning to consultants with deep domain expertise to help them transform specific business processes.
Hello again, 1998
A couple of years ago, AI startups moved away from the cloud-only models of SaaS and started deploying on-prem again. Deploying AI onsite was simply more palatable to customers than deploying via cloud alone. Hello again, 1998.
But something else happens when you do custom deployments onsite: customers expect you know what you're talking about.
In the SaaS world, customers bought your software. They sometimes took your advice about how to use the software, but mostly they designed their own workflows. But in the AI world, no one has any patterns to follow about how people and AI collaborate, how workflows get redesigned, when to build vs. buy, or anything else.
In this context, customers expect their AI vendors to be consultants, and they expect their consultants to be AI vendors. The composition of both technology companies and consultancies is changing accordingly.
So far in 2026, I have connected with founders of nearly 100 B2B startups. Once startups pass about 10 employees, nearly all have at least one Forward Deployed Engineer (FDE) on staff. But a growing number of them (31 of the 98) also have explicitly consultative roles like Customer Solutions Architect, AI Transformation Consultant, and Forward Deployed Strategist.
Eight months ago when I looked at startup team composition, these explicitly consulting-oriented roles were simply not present. The patterns of AI team composition have not yet stabilized because the work itself is still rapidly changing.
I wonder what Act 4 will bring.
Kieran
If you liked this story, why not subscribe to nerd processor and get the back issues?
My latest data stories | Build like a founder | nerdprocessor.com