AI won’t fix how your company works.
Blueprint the business before you automate it.
Eight months ago, I did an AI value chain analysis and never posted it. Since then, we’ve seen infrastructure constraints, pressure on chips and energy, rising implementation costs and, most importantly, a growing gap between AI adoption and actual business value.
Yet most of the AI conversation I see still focuses on where this is all going…doom and gloom or land of prosperity.
I’m more interested in a practical question: how do you put AI into an existing business and make it work?
Most companies have fragmented data, disconnected processes, undocumented institutional knowledge, legacy technology and people already dealing with significant change. That’s the environment AI has to work within.
From the work we’re doing, four things stand out.
Blueprint the business before you automate it
McKinsey found that 88% of organizations are using AI in at least one function, yet only 39% report enterprise-level EBIT impact. Workflow redesign is one of the factors most strongly associated with meaningful impact.
88%
of organizations are using AI in at least one function
39%
report enterprise-level EBIT impact
Poor data, broken processes and unclear decision rights don’t disappear when you add AI. You can automate inefficient work, accelerate bad decisions and create new bottlenecks further down the workflow.
There’s another issue emerging: every function is building its own AI agenda. Marketing has a roadmap. Finance has priorities. Customer Care is exploring agents. HR is looking at productivity. IT is trying to govern all of it.
The problem is that work doesn’t follow the org chart.
Start by understanding the end-to-end workflows that create value across the business. Map the people, processes, technology, data, decisions and friction around them. Then determine what should disappear, what needs redesigning, where AI can help and where human expertise still matters.
Otherwise, you risk optimizing individual functions while making the overall system worse.
The economics need to work
The cost of accessing AI intelligence has fallen dramatically, but enterprise implementation has a much bigger cost base.
Model usage, compute, integration, data readiness, governance, training, maintenance and human oversight all need to be considered. Usage costs that look insignificant in a pilot can also become meaningful at enterprise scale.
A good AI business case should answer three questions:
What will this really cost? What measurable value will it create? What work or cost are we actually removing?
If those answers aren’t clear, productivity gains can quickly become another layer of technology spend.
Design the human + agent operating model
Agents introduce a much bigger organizational question than “human in the loop.”
Companies need to decide what AI can assist, recommend, automate or own, where people remain accountable, and how work moves between the two.
Roles will change. Skills will change. Managers may oversee combinations of people and agents. Performance measures, career paths, training and ways of working will need to evolve with them.
Culture matters too. Employees who see AI primarily as a headcount reduction exercise will respond very differently from employees who understand how their work is changing and where their judgment, creativity and relationships become more valuable.
These decisions need to be designed into the transformation from the beginning.
Precision first. Scale second.
There is pressure to move quickly, and waiting carries risk. But scale should follow evidence.
Start with use cases tied to clear business outcomes. Understand the economics. Measure the value. Fix the data, process and organizational issues you uncover. Then expand while building the governance and capabilities required to support it.
For CEOs and commercial leaders, I’d start with five questions:
- Where does our organization create value, and where could AI materially improve it?
- Are we fixing the workflow before we automate it?
- Are functions pursuing AI independently, or are we looking across the enterprise?
- What are the true economics, including what work or cost will actually disappear?
- How will roles, accountability and culture change as people and agents work together?
AI adoption is easy to measure. Business transformation is much harder.
The organizations that get this right will understand the business they are putting AI into, redesign the work where necessary, and scale based on measurable value.
