AI made the work faster. It also broke the pricing.

Jonas Dahlberg has spent the past year close to enterprise AI decisions, and one pattern stays with him: how much still must change on both sides of the table. Companies have to rebuild the way they work, and the firms selling them AI have to rethink how they price it.  

The dependency runs deep. Customers increasingly lean on external consultants for the AI work that matters most, and those consultants are using AI to build it. Whose AI, running on whose data, paid for by whom, are no longer small questions.

On either side of the table, few have made that shift. And when the work moves this fast, the old way of pricing and contracting it no longer holds.

Rebuilding core processes around AI makes it faster

The distinction Jonas keeps coming back to is between layering AI on existing processes and rethinking them.  

His advice to clients is the one most firms are slow to give:

"Don't try to take your current process and make it agentic. Think about what the input is and what the outcome should be. Then figure out how to rebuild that with AI at the centre"

He points to a recent engagement with an enterprise legal function reviewing thousands of contracts a year across dozens of legal entities and languages. The journey from a customer wanting to buy something to a signed contract could take weeks of manual review and escalation. Adequately staffed, well-intentioned, but designed for a different era.

The instinct in most organizations is to find a tool that reads contracts faster. Jonas sees that as the wrong starting point. What changed the outcome was building the solution around the organization's own policies, jurisdiction-specific rules, and escalation paths, so it focused on the outcome the customer wanted and not necessarily the way that business already worked. Review time then naturally fell by 70 percent.

"Get the lead time down and it makes a significant difference to the top line. That's what rebuilding the process with AI as the engine looks like."

Speed changes the economics

The payoff of rebuilding around the core process is speed. That speed creates a problem the industry has not settled. When AI does the heavy lifting, the hours that used to define the work stop measuring its value, and the cost of producing it shifts too.

"One hour of a person working with AI can be worth many times what that hour used to be. Paying by the hour stops making sense. You pay for the outcome, and the cost of running the AI is part of producing it, and of operating it once it is in production."

That last part is where most arrangements fall silent. Running AI at production scale costs real money in compute, and someone has to carry the cost. Jonas's view is that the party getting the value should be the one paying for it, stated openly rather than buried in a rate.

"Compute is a real cost now. The honest thing is to put it on the table. Who is using the AI, who is paying for it, and what the client is actually buying. Leaving that in a grey zone serves no one."

The contracts have not caught up

The same silence shows up in how the work is governed. When a consultant runs client data through an AI tool, the basic questions often go unasked: which tool is it, has procurement reviewed it, does the NDA even cover it.

Most agreements were written for a world where none of this existed. The tool, the data handling, the compute cost, and what the client is actually paying for all sit in a grey zone that suits no one.

For Redeploy, this has led to a deliberate shift in how contracts are written, with explicit clauses on which AI tools are in use, how client data is handled, and what the client is paying for. Not because it is legally required yet. Because ambiguity does not serve anyone.

It is an early signal of what AI-first actually demands. The firms that pull ahead over the next 12 to 18 months will be the ones that rebuild the core process for real, move at the speed it unlocks, and stay honest about what it costs and who carries it.