Most business have at least one AI programme running, but many of them stall in the same place: a long list of pilots, a thin spread of budget, and no single domain transformed deeply enough to change the P&L.
A study of 20 AI-leading companies showed that the ones generating returns focused their efforts on one to three business domains rather than papering AI across the organisation. This guide sets out how a mid-market C-suite can identify those domains, sequence the work, and turn AI investment into game-changing value.
What is an economic leverage point?
An economic leverage point is the part of the business model where a small improvement produces an outsized financial result. Every industry has them, and they are usually obvious once the question is asked.
- Leverage points are domain-level, not function-level. Forecasting and planning in retail, claims processing in insurance, yield in mining, supply chain integration in automotive: these are end-to-end domains rather than tasks inside a single team.
- Leverage points map to the P&L directly. A 2% improvement in process yield or a 5-day reduction in cash conversion shows up in EBITDA in a way that a faster chatbot response time does not.
- Most companies have three or fewer. The instinct is to find ten; the discipline is to find three proper ones.
How do we identify our leverage points?
Ask yourself these questions:
- Which three line items, if improved by 10%, change next year's earnings the most? Run the numbers across revenue, cost of sales, and operating expenses; the candidates fall out of the spreadsheet, rather than the strategy deck.
- Where does the business already lose money to bad decisions? Pricing leakage, churn, returns, downtime, fraud, write-offs, and stockouts are usually where data and AI produce the fastest gains because the decisions repeat at volume.
- Which workflows cross the most functions? The domains that span sales, operations, and finance tend to carry the highest leverage because integration is where mid-market companies lose the most time and margin.
How do we sequence the work?
Concentration only pays off if the foundational capabilities are built in the right order.
- Define the domain and the financial target. Name the business outcome (gross margin, days sales outstanding, customer retention), the size of the prize, and the timeframe; this is the contract between the C-suite and the programme.
- Fix the data foundation in that domain. Clean, governed, business-consumable data inside the chosen domain matters more than an enterprise data strategy that touches everything and finishes nothing.
- Redesign the workflow around the decision. An AI prediction creates value only when the upstream and downstream processes change to act on it; the redesign is the work, the model is the input.
- Deploy the AI capability with guardrails from day one. Trust, testing, and human-in-the-loop controls are the difference between a deployable system and a demo.
- Measure against the financial target, monthly. Cash-accretive AI programmes show movement inside 12 to 24 months; the absence of movement is a signal to redesign.
What capabilities does a mid-market C-suite need to build?
The same capabilities the global leaders build, scaled to mid-market reality.
- A leadership team that has done a learning journey. Conviction at the top is the precondition for every other decision; without it the programme drifts back into IT.
- A small, high-density team inside the business. A few competent builder-engineers embedded in the domain outperform a large team of generalists, and they are easier to hire and retain at mid-market scale.
- A reusable data and AI platform. Build once, reuse across the chosen domains; reuse rate, cost-per-call, and time-to-onboard are the metrics worth tracking.
- Domain ownership at the executive level. A named C-suite owner per domain, accountable for the financial outcome, with the authority to redesign the workflow.
- A trust and risk function from the start. Model testing, data lineage, access controls, and red-team review are foundational capabilities, rather than a compliance afterthought.
How does the C-suite split the work?
Each role carries a defined part of the programme. Ambiguity at this level is the most common failure mode.
- The CEO sponsors the focus decision. Naming three domains and protecting that focus against the pull of every other priority is the work only the CEO can do.
- The CFO owns the value case. Sizing the prize, funding the foundational work, and measuring the financial result keeps the programme honest.
- The COO redesigns the workflow. Domain ownership and process redesign sit naturally with operations; the AI model is one input into a re-engineered way of working.
- The CIO or CTO builds the platform. Reusable data, reusable services, and the security perimeter around them are the technology agenda, rather than the use-case agenda.
What does success look like in 24 months?
- One to two years to breakeven. Cash accretion is achievable inside the first cycle when the focus is on target and the foundational work is laid out.
- 20% EBITDA uplift in the transformed domains. The gains come from depth in a few areas, rather than breadth across many.
- Three dollars of incremental EBITDA for every dollar invested. The return curve rewards concentration and capability-building over tool-buying.
Where can mid-market programmes go wrong?
We see these patterns occur:
- Spreading the budget across every function. Every team gets a little, nothing gets enough; the programme produces motion without movement.
- Treating AI as an IT initiative. Delegating ownership to the technology team removes the business accountability that helps the team adopt it.
- Buying tools before building capabilities. Platforms, data products, talent density, and workflow redesign should not be skipped; the companies that try to, end up rebuilding the foundation later at higher cost.
The takeaway
Focus determines AI advantage more than the choice of technology does. Mid-market companies that pick three economic leverage points, build the capabilities those domains depend on, and hold the focus at C-suite level are the ones generating value in the business.
This article draws on themes from McKinsey's AI transformation manifesto (April 2026) and the second edition of Rewired by Lamarre, Smaje, Singla, Sukharevsky, and Levin.