AI productivity is often framed as a time-saving exercise. For executives, the bigger opportunity is to identify where AI can reduce costs, release capacity, improve outcomes and make valuable new work possible.

A useful way to frame the decision is through two tools: a stopwatch and a map.

A stopwatch tells you whether AI makes the organisation faster. A map asks whether it enables the organisation to go somewhere previously out of reach.

The stopwatch measures existing work:

• How quickly can we prepare this report?
• Can we respond to customers sooner?
• Can we reduce the effort required for routine analysis?

The map considers a different operating model:

• Which customers could receive useful analysis who currently receive none?
• Which ideas could we test before making a significant investment?
• What could employees attempt with support beyond their usual skills?
• Which risks or opportunities could we investigate more consistently?

A time-saving business case evaluates the current organisation. An opportunity business case considers a newly feasible version of it.

However, every map needs a destination. Producing more reports or research has limited value by itself. A stronger proposition connects the capability to a business result:

We can now analyse neglected customer accounts and test whether targeted engagement improves profitable sales.

That is specific enough to evaluate.

Look beyond the visible backlog

Every organisation has a visible backlog: approved work waiting to be completed.

It also has an invisible backlog: potentially valuable work people have stopped proposing because time, cost or specialist capability made it impractical.

The visible backlog

• Reports awaiting completion
• Customer enquiries awaiting replies
• Projects waiting for a specialist
• Scheduled reviews running late

The invisible backlog

• Questions nobody has capacity to investigate
• Customers receiving little proactive contact
• Ideas abandoned before expert review
• Reviews considered too expensive to undertake

The invisible backlog can contain some of the strongest opportunities for AI transformation.

Executives can begin with a simple question:

What do our people regularly say they would do if they had more time or access to specialist expertise?

Some suggestions will have limited value. Others may reveal opportunities for revenue, resilience, customer service or better decision-making.

AI changes the cost of attempting the work. The organisation still needs to establish whether the result deserves further investment.

Saved time is capacity

When AI releases staff time, four outcomes are possible:

1. Expenditure falls through reduced overtime, contractor use or another avoidable cost.
2. The same team completes more valuable work.
3. Work becomes more sustainable, with less pressure and after-hours activity.
4. The capacity is absorbed by low-value activity.

These outcomes should be measured separately.

A randomised study involving 7,137 knowledge workers across 66 firms estimated that employees encouraged to use generative AI spent around two fewer hours each week on email during the latter part of the experiment. They also completed less work outside regular hours.

The researchers found no detected change in the quantity or composition of tasks resulting from individual access to the tool. The study therefore shows time being released without establishing additional business output.

Saved time is capacity. What the organisation does with it determines the return.

Executives evaluating AI productivity should ask:

If this team gets five hours back each week, what do we want to happen next?

Without a clear answer, estimated time savings remain potential capacity rather than captured business value.

Lower unit costs can increase total spending

AI can make each piece of work cheaper while increasing total expenditure because more activity becomes economically feasible.

Consider a hypothetical customer-review programme:

• Ten traditional reviews cost $1,000 each: $10,000 in total.
• One hundred AI-assisted reviews cost $200 each: $20,000 in total.

The cost per review falls by 80 per cent, while total spending doubles.

This could be a strong investment if the additional coverage improves retention, identifies sales opportunities or reduces risk. It could also create a greater volume of work with limited commercial effect.

The programme has expanded, so the expansion should be evaluated on its own merits.

Comparing the new programme with the theoretical cost of 100 traditional reviews would create an artificial saving if the organisation never intended to purchase that volume.

The executive question becomes:

What would we choose to do at greater scale if its cost fell—and how would we recognise the point of diminishing returns?

Use GAIN to define the business value

Efficiency and opportunity explain how AI can help. An AI business case also needs to show how the organisation expects to capture value.

The GAIN framework provides four routes:

Generate savings

What expenditure disappears?

This is the clearest financial route. It includes spending that disappears, such as paid overtime, avoidable contractor costs or duplicated services.

Time saved by salaried employees should only be treated as a financial saving when it leads to an actual reduction in expenditure.

Apply capacity

Where will released time create value?

Released time can allow the same team to serve more demand, reduce queues or complete work that had been repeatedly deferred.

The business case should name the work that will fill the capacity. “Employees will have more time” is an operational observation. “Employees will use that time to review an additional 20 high-value accounts each month” is a testable proposition.

Improve outcomes

What becomes better?

AI may help an employee produce better work within a similar period.

The published customer-support study found an average 15 per cent increase in issues resolved per hour, with benefits varying considerably by worker experience and skill. The researchers measured a completed service outcome rather than text-generation speed alone and cautioned against broad generalisation from one company, occupation and tool.

Improved value may appear through:

• Greater accuracy
• More consistent service
• Faster resolution
• Fewer errors
• Reduced rework
• Stronger analysis or recommendations

New possibilities

What valuable activity becomes feasible?

AI can make previously impractical activities feasible.

In a field experiment at P&G, individuals using AI achieved performance comparable to teams without AI on the product-innovation tasks studied. Their proposals also combined commercial and technical perspectives more evenly.

This supports the possibility that AI can broaden employee capability. It does not mean every employee can replace a specialist team or that every AI-supported idea will succeed.

AI creates potential. GAIN shows how the business intends to capture it.

Keep financial and operational scorecards separate

Executives need visibility across two scorecards.

Financial results

• Realised cost reductions
• Additional contribution
• Revenue influenced
• Full implementation and operating costs

Capability and operating results

• Capacity released
• Work completed
• Service coverage
• Quality and rework
• Opportunities being tested
• Employee sustainability

Operational improvements should remain visible even when a credible financial conversion is unavailable. At the same time, capability measures should eventually connect to useful organisational outcomes.

Care is also required to prevent double counting. If released employee capacity supports additional sales, counting its full salary-equivalent value as a saving alongside the resulting contribution would overstate the return.

AI can move the constraint

Imagine AI could produce every sales proposal your organisation wanted by tomorrow.

Would sales increase, or would managers simply receive more proposals to review?

Faster production creates limited value when approval, customer demand or delivery capacity remains the real constraint. AI may move the bottleneck to another part of the process.

For each proposed use case, ask:

• What currently constrains the complete process?
• Who will review the additional output?
• Can the organisation act on it?
• What work will review and follow-through displace?
• How will quality and rework be measured?

The quality check is essential.

In the BCG experiment, AI improved performance on tasks suited to its capabilities. On one task beyond those capabilities, participants using AI were 19 percentage points less likely to reach the correct solution.

The result is task-specific, yet it reinforces a practical principle: measure end-to-end performance, including verification and rework, rather than draft speed alone.

Run experiments that buy evidence

AI can make it cheaper to determine whether an idea is promising. It cannot make every idea successful.

For new activities, begin with staged questions:

1. Can we produce something reliable enough to use?
2. Does somebody use it to make a decision or take action?
3. Does that action improve the outcome?
4. Is the improvement worth the complete cost?

Define the decision each experiment will inform before it begins. A successful pilot may support expansion, modification or an evidence-based decision to stop.

All three outcomes can protect capital and management attention.

Applying GAIN with Theta Assist

A balanced proof-of-value exercise can test two use cases:

• One existing task: measure end-to-end effort and quality before and after AI assistance.
• One previously neglected activity: test whether the new capability produces a useful action and measurable outcome.

The evaluation should include setup, AI usage, employee preparation, review and follow-through. Existing salaries may remain unchanged, but employee attention still carries an alternative use.

Theta Assist provides a business-customisable marketplace of AI assistants, supported by onboarding and ongoing assistance. This creates a foundation for repeatable, governed business uses across teams.

Theta can help organisations identify where AI could:

• Generate genuine savings
• Apply released capacity to valuable work
• Improve operating and customer outcomes
• Enable new possibilities

The objective is to test each route to value rather than assume the return will appear.

Start with the work left undone

The executive conversation around AI should extend beyond salary savings.

Ask two questions:

1. Can we perform today’s work more efficiently or effectively?
2. What worthwhile work could we undertake that currently remains untouched?

The first needs a stopwatch. The second needs a map.

Saved time needs a useful destination. New capability needs evidence of value.

Broaden the definition of AI value while maintaining the standard of proof.

Book a demo to explore where your organisation could GAIN value from AI.

References

• Brynjolfsson, E., Li, D. and Raymond, L. R. Generative AI at Work. The Quarterly Journal of Economics.
• Dell’Acqua, F. et al. Navigating the Jagged Technological Frontier. Organization Science.
• Dillon, E. W. et al. Shifting Work Patterns with Generative AI. American Economic Association.
• Dell’Acqua, F. et al. The Cybernetic Teammate. National Bureau of Economic Research.
• Humlum, A. and Vestergaard, E. Still Waters, Rapid Currents. Chicago Booth.