Deferred, reframed, or dropped
Everyone writes about what it takes to adopt AI: culture, leadership buy-in, change management, and training programs. It is always the same list, and it’s almost entirely useless. That is not because it’s wrong, but because the painful parts keep getting treated as optional.
Across most software organizations I work with, the comfortable parts get done (the announcement, the pilot, the working group, the vendor evaluation). But the painful parts get deferred, reframed, or quietly dropped. And then people wonder why the numbers didn’t move.
Why?
The painful parts come down to the same three people/functions, each of whom has to give up something.
The P&L owner
Not a sponsor but the person who owns the number AI is supposed to move. Adoption always shows up as cost first, and when the quarter gets tight, that cost needs someone who feels the pain of not doing it. Without them, it dies quietly in a budget review, usually described as a pause.
The roadmap owner
Adoption that sits next to the roadmap is mostly for show. Until something planned gets canceled to make room, nothing has actually been decided.
The developers
Most adoption advice goes straight to tools and mandates, and people wonder why nothing sticks. When tools are imposed on workflows that developers understand better than anyone, they find ways around them without ever pushing back directly. The tool gets used on paper, but in reality nothing changes.
Gamechangers
There are always one or two people who are already a few steps ahead, experimenting on their own, figuring out what works, becoming the go-to people when their colleagues get stuck. And that is what actually shifts a team. Knowledge spreads sideways faster than any training program can accomplish. Find these go-to, curious, adventurous people, give them room to explore and experiment, and get out of the way.
So what should you do?
The costs are almost always higher than the forecast, and almost always in the same three places.
You have to pull the right people out of delivery
AI Adoption that is run as a side project produces side results. The single decision most organizations refuse to make is to take their best people off billable work. And that is why most AI adoption ends up staffed by whoever happens to be free.
You have to budget past the pilot
The pilot is the cheap part, but the integration and the rework -the second and third attempt after the first approach fails- those add to the real number and sit well above the business case. Pretending otherwise means the effort runs out of money halfway through and gets quietly shelved.
It may hit your revenue before it helps it
This is the one that catches most companies off guard and is especially applicable for your services. If you bill by effort and the work now takes half the time, you are doing the same work for less money. The revenues will drop before you have repriced your model. The margin drop arrives before the efficiency gain does.
The use case
Under all of it is the question that should have come first and almost never does: is there even a use case?
Not a demo and not a use case that sounds good in a vendor meeting, a real workflow, one with enough volume, enough structure, and enough revenue attached that automating it moves a number that someone is accountable for.
Most adoption efforts never find one, and spend eighteen months discovering that. Yet not moving forward poses an even greater risk. It will eliminate the moat you have carefully built around your company and open the playing field for competition from all angles.
By Bogdan Robu
Delivery Manager @ Yonder
Banner by Unsplash, Roman Budnikov
STAY TUNED
Subscribe to our newsletter today and get regular updates on customer cases, blog posts, best practices and events.
