Every piece of software that runs your organization—your CRM, your ERP, your project management tool, your productivity suite—was built on one assumption: the primary user was a human. That assumption is no longer true.
Agents are already in your software stack, at machine speed, with no menus or training, doing work in seconds or minutes that used to be done by humans in hours or days. The question is what does that change and for whom.
The user has changed, and that changes everything
Every enterprise piece of software for 30 years was built around one constraint: The human on the other end. Each feature had to be discoverable. All workflows had to be teachable. The limits of software were bounded by what a person could navigate. That ceiling is gone when the agents are doing the work. Software no longer needs to be powerful, or useful.
You can see that change already in evidence. One of our finance managers within one of our commercial finance teams recently described sitting down with a messy data set, a blank workbook and one goal: get to the business story faster. They asked Copilot to pivot from a raw data pull to the product level. When the first pass came back as a full-year view, they typed back in plain English that they needed fiscal quarters. Copilot navigated back into the source, added a column mapping each row to the correct quarter, and rebuilt the pivot. The manager kept on asking for a Volume-Rate-Total decomposition and a year on year contribution bridge. And finally, to stress test the output, they purposely broke a few formulas and asked Copilot to audit the entire workbook. It walked every tab, flagged every inconsistency with a severity level, rewrote every broken formula, all while the manager answered emails and took calls on a second screen. Their reflection captures the shift: “It works quietly in the background, while we can focus on insights, decision-making and meaningful business-partner conversations.” The manager was not running excel. AI was.
That bit about “quietly in the background” is important. Many of the most important players won’t even appear as a chat window. They’ll run headless, triggered by a policy change, a data refresh, a ticket being opened, a shipment being delayed, doing work inside systems at machine speed, then surfacing results only when a person needs to review, approve or step in.
Software is being redesigned for agents at three layers
Agents are already working inside your tools. The platforms that pull ahead will be transformed from the data up.
What this means for your existing software
- A lot of the discussion around AI and enterprise software is skin deep – the interface, the features, the speed. And that’s all important, but more consequential changes are taking place under the hood.
- User experience. It’s not a view that interfaces will disappear entirely as agents take over, but tech doesn’t propagate like that. Adoption is about meeting users where they live – in the tools they know and the canvases where work lives. Interfaces are where work is reviewed, shared and handed off. Interfaces are the rendezvous point. Now there are two kinds of user, human and agent, and software has to serve both.
- Business rules. The layer that encodes how a company works – how you close the books, how a report gets approved, who a case gets escalated to. Today, that logic is embedded in workflows built for humans. The more execution is done by agents, the more it needs to be built into the system as skills that an agent can directly invoke. That’s where the big efficiency gains will come from.
- Preparing data. Every enterprise application uses data as the basis for its operation, but agents gain from this data being optimized for their use. Agents can learn to understand the structure and meaning of a dataset by themselves — but if they have to do that every time someone asks a question the AI has to reinvent the wheel over and over again. The answer is to process the data as it comes in so the agents can immediately begin answering the question rather than trying to figure out what it is they are looking at. Imagine it like the difference between handing someone a big stack of papers and a well-organized brief. Same info, but a very different starting point.
Implications for your organisation
As agents are taking on more and more of the execution and the barriers to creating software are lower than ever, a reasonable question comes up: why not just build your own? AI lets you build your own CRM, but every hour you spend building and maintaining software that could be done by a ready-made solution is an hour you’re not spending on the work that defines your competitive edge. The AI era will demand a reckoning on how organizations are spending their time and money. The companies that will lead are not the ones that do the most, they are the ones that are most disciplined about what only they can do. We are seeing less insourcing, more concentration.”
The same goes for the tools already owned or subscribed to by your organization. Companies pay for software with tons of features that nobody or hardly anybody uses. Agents will discover and use those features. And they may even begin to ask for capabilities that no human user could have imagined. Software is going to change faster than any one human can keep pace with, and that makes the human layer more important, not less.
As human work shifts upstream – less hands-on time in the software, more time deciding what it should produce – the organizations that pull ahead will be those that deliberately develop their employees’ ability to set direction, evaluate outcomes, and keep accountable for how the system performs. That’s a talent and culture investment, not a technology investment.” And the time to start investing is today.
