When I speak with companies about their AI journey, I often see what they call AI strategy today as a list of activities, not a plan for competitive advantage. Let’s break this down.

 

Activity is being mistaken for advantage.

 

Look closely at most AI strategy documents, and you will find the same ingredients: the same tool rollouts and licences, a handful of pilots, some prompt training for staff, maybe a governance policy. This is all useful, but none of it is a strategy.

A strategy answers the question: how does this change our position relative to competitors? An activity list answers a different question: what have we done? Leadership teams often confuse the two, because activity is easy to show and advantage is hard to prove.

 

Here’s a test worth running:

If your closest competitor rolled out the exact same tools tomorrow, would you still be ahead?

 

The technology itself has become widely available. When the tool is available to everyone, the tool cannot be the source of your advantage. It never was, in any technology wave.

What decides who wins is what a company builds around the tool. Most AI strategies stop short of this. They focus on access — who has the licence, who completed the training, how many people log in each week — instead of focusing on change: how work gets done, how decisions get made, how customers experience the business, how fast the organisation can move.

Adoption numbers get mistaken for progress too. A high usage rate tells you people are using AI. It does not tell you whether the company is becoming more competitive. I have seen organisations with excellent adoption figures and no measurable change in cost, cycle time, quality, or revenue. Usage is a leading indicator at best. On its own, it proves nothing.

 

 

 

Where the advantage actually comes from

 

1. Your proprietary context

The model everyone can buy is generic. Your internal knowledge, your customer data, your decision rules, the way your best people handle exceptions — none of that is available to your competitors. Combine commodity AI with knowledge that is genuinely yours, and the output stops being commodity. This is the biggest lever most companies leave unused, because capturing that knowledge properly takes more effort than switching a tool on.

 

2. Workflow redesign, not workflow addition

Bolting AI onto an existing process usually produces a small, useful improvement. Someone drafts faster, someone summarises quicker. Worth having, but rarely a shift in competitive position, because competitors can copy a small improvement in an afternoon. The advantage appears when a company redesigns the workflow around what people do well and what AI does well, including who has authority to decide what. That kind of redesign is much harder to spot from the outside, and much harder to copy quickly.

 

3. Speed of learning

Some organisations can identify a promising use case, test it, measure whether it works, and either kill it or scale it within weeks. Others take two quarters to get a single pilot through committee. That gap compounds over time. It isn’t really an AI advantage at that point — it’s an organisational one that AI simply makes visible faster.

 

4. Governance built for speed, not caution

Clear decision rights and clearly defined risk boundaries let teams move faster, because everyone knows what they can decide themselves and what needs review. Heavy approval layers do the opposite. They protect the company from small risks while costing it the one advantage AI was supposed to deliver: speed. Good governance is an accelerator. Bureaucratic governance cancels out the advantage before it appears.

If adoption and licence numbers are the wrong scoreboard, the right one is business outcomes measured against a baseline you can defend:

• Cycle time for a defined process, before and after
• Cost per unit of output for a specific workflow
• Conversion or win rate where AI touches the customer journey
• Quality or error rate in a process that used to rely on manual review
• Revenue attributable to a capability that did not exist before
• Risk exposure reduced in a process that used to depend entirely on individual judgement

None of this requires an exotic measurement system. It requires picking a handful of workflows, knowing where they stood before you touched them, and tracking where they stand now.

 

 

What companies should do now

 

• Step 1: Run the copy test on your current roadmap. For each major initiative, ask honestly whether a competitor buying the same tools tomorrow would erase the benefit.
• Step 2: Map your proprietary context. Identify the knowledge, data, and decision rules that are genuinely yours, and prioritise use cases built on them rather than generic tasks anyone could automate.
• Step 3: Redesign one workflow properly. Rather than adding AI to an existing process, rebuild it around what people and AI each do best, including who has authority to decide what.
• Step 4: Set a baseline before you scale anything. Choose two or three workflows that matter most and record cycle time, cost, quality, or conversion before you change them.
• Step 5:  Check whether governance is helping or slowing you down. If approvals take longer than the work itself, the structure is costing you the speed advantage AI was meant to create.
• Step 6: Replace vanity metrics in your reporting. Swap usage percentages and training completion rates for the outcome measures above, at least in what goes to leadership.

 

AI itself is becoming a commodity. Every competitor can buy the same models you can. What cannot be bought is the system you build around it — your workflows, your data, your expertise, and your ability to learn and execute faster than the company next to you. That system is where the advantage actually lives, and it is the part most AI strategies still leave out.