How to Build an AI Business Case: From AI Idea to Measurable Business Value

AI investment is becoming easier to approve — and harder to justify.

A new AI tool promises productivity gains.

A business function identifies an opportunity to automate work.

A vendor demonstrates an impressive capability.

A competitor announces an AI initiative.

A leadership team decides it needs to “do more with AI”.

The temptation is to move quickly.

But before asking which AI technology should we buy?, leadership teams should ask a more fundamental question:

What business outcome are we investing to create?

That is the foundation of an effective AI business case.

Because AI does not create value simply because it is deployed.

Value is created when AI changes a business outcome enough to justify the investment required to achieve it.

What is an AI business case?

An AI business case is a structured assessment of whether a proposed AI investment is likely to create sufficient business value to justify its cost, complexity and risk.

A strong AI business case should explain:

• the business problem or opportunity;

• the outcome the organisation wants to change;

• why AI is an appropriate intervention;

• the expected financial and non-financial benefits;

• the total cost of implementation and operation;

• the organisational changes required;

• the risks and dependencies;

• how success will be measured;

• who owns the outcome;

• what evidence would justify scaling or stopping the investment.

• 
  

This is different from a technology proposal.

A technology proposal explains what the AI can do.

A business case explains why the organisation should pay for it.

That distinction is critical.

Why traditional AI business cases often fail

Many AI initiatives begin with an assumption:

“AI will make us more productive.”

That may be true.

But productivity is not automatically financial value.

Suppose an AI tool saves an employee 30 minutes every day.

Across 1,000 employees, that sounds substantial.

But several questions remain.

Did employees actually save the time?

Did they use the technology consistently?

What did they do with the time saved?

Did the organisation increase output?

Did service improve?

Did costs fall?

Was additional hiring avoided?

Did revenue increase?

If nothing else changed, the organisation may have created theoretical productivity without creating measurable commercial value.

That is why the business case needs to establish the path from AI capability to business outcome before the investment is made.

Start with the business problem, not the AI

A weak AI business case might begin:

“We want to deploy generative AI across Customer Service.”

A stronger one begins:

“Customer service advisers spend approximately 25% of their time searching multiple knowledge sources for information, increasing response times and reducing capacity.”

Now there is something to evaluate.

The leadership team can ask:

How significant is the problem?

What does it currently cost us?

What would improvement be worth?

Is AI the best way to solve it?

This prevents an increasingly common mistake:

starting with an AI solution and then looking for a problem to justify it.

Sometimes AI will be the right intervention.

Sometimes conventional automation, integration, process redesign, better data or simply removing an unnecessary process will create more value.

An independent AI business case should be capable of concluding:

Don't invest.

The Stratify AI Value Chain

For an AI investment to create measurable value, several things normally need to happen.

BUSINESS PROBLEM

↓

AI OPPORTUNITY

↓

INVESTMENT

↓

IMPLEMENTATION

↓

ADOPTION

↓

BUSINESS OUTCOME

↓

FINANCIAL VALUE

Each stage matters.

An attractive opportunity without successful implementation creates no value.

Successful implementation without adoption creates little value.

Adoption without changed business outcomes creates activity, not necessarily value.

And improved business outcomes that cannot be translated into commercial impact may still make the financial return difficult to establish.

This is why AI ROI should not be treated as a Finance calculation performed at the end of a project.

Value needs to be designed into the initiative from the beginning.

Step 1: Define the business problem

Before estimating benefits, define the current state.

What is happening today?

How frequently?

How many people are affected?

How much time does it consume?

What does that cost?

What impact does it have on customers?

What errors occur?

What delays are created?

What revenue opportunity is being missed?

The more clearly the current state can be quantified, the easier it becomes to measure improvement later.

For example:

“AI could improve invoice processing.”

is vague.

Whereas:

“The finance team manually processes 12,000 invoices each month, requiring approximately 900 hours of employee time and generating an average of 180 exceptions requiring additional investigation.”

creates a baseline.

Now an investment can be evaluated against something real.

Step 2: Define the outcome

Next ask:

What would need to change for this investment to be successful?

Possible outcomes include:

Revenue

AI may increase:

• sales conversion;

• customer retention;

• average order value;

• cross-selling;

• pricing effectiveness;

• market reach;

• sales capacity.

Cost

AI may reduce:

• manual processing;

• external expenditure;

• error correction;

• administration;

• service costs;

• recruitment requirements.

  
  

Capacity

AI may allow existing employees to handle:

• more customers;

• more transactions;

• more analysis;

• more cases;

• more projects;

  
  

without equivalent increases in headcount.

Speed

AI may reduce:

• response times;

• processing times;

• research time;

• reporting cycles;

• product-development time.

  
  

Quality

AI may improve:

• consistency;

• accuracy;

• customer experience;

• decision quality;

• compliance;

• information availability.

  
  

Not every benefit needs to become a direct cost saving.

But leadership should understand which outcome is expected to change and why that outcome matters commercially.

Step 3: Establish the baseline

This is one of the most frequently missed steps in AI value measurement.

If you don't know what performance looked like before AI, proving improvement afterwards becomes difficult.

Before implementation, record the relevant baseline.

Depending on the use case, that could include:

Cost per transaction

Average handling time

Revenue per employee

Conversion rate

Customer response time

Error rate

Employee hours required

Customer satisfaction

Time to produce an output

Number of cases completed

Then compare performance after implementation.

Otherwise organisations risk reaching the end of an AI project and asking:

“Do we think it helped?”

That's not benefits realisation.

That's opinion.

Step 4: Calculate the total cost of AI

The price of the AI software is not the total cost of the AI investment.

A credible business case should consider the total cost of ownership.

That may include:

• software licences;

• model/API usage;

• implementation;

• integration;

• data preparation;

• infrastructure;

• security;

• consultancy;

• employee training;

• change management;

• governance;

• monitoring;

• support;

• ongoing maintenance;

• internal employee time.

  
  

There can also be opportunity costs.

If your best technology and transformation people spend six months implementing one AI initiative, what else aren't they doing?

Leadership needs to understand the full investment, not merely the vendor invoice.

Step 5: Account for adoption

This is where many theoretical AI business cases become unrealistic.

Suppose the model assumes:

2 hours saved per employee per week × 1,000 employees.

That benefit assumes employees actually use the technology effectively.

What if only 40% adopt it?

What if adoption varies significantly between teams?

What if employees use it, but only for tasks unrelated to the expected benefit?

The business case should therefore incorporate realistic adoption assumptions.

For example:

Potential benefit at 100% adoption: £1 million

Expected adoption in Year 1: 60%

Realistically addressable benefit: £600,000

Then apply further assumptions about how much of that benefit can genuinely be realised.

This is why workforce readiness and ROI cannot be separated.

Read Why Employees Aren't Adopting AI: 7 Barriers Business Leaders Need to Address 

Step 6: Distinguish productivity from realised value

This distinction could become one of the central principles of The AI Revenue Machine.

Consider three levels:

Level 1 — Time saved

AI reduces the amount of time required to complete a task.

Level 2 — Capacity created

The organisation redirects that time towards additional productive activity.

Level 3 — Value realised

That additional capacity produces a measurable commercial outcome.

For example:

AI saves a salesperson five hours per week.

That's time saved.

The salesperson uses those five hours for additional customer conversations.

That's capacity created.

Those conversations generate £40,000 of additional annual gross profit.

That's value realised.

Those are not the same thing.

Leadership teams should therefore ask:

What will we do with the capacity AI creates?

Without an answer, productivity claims can become significantly overstated.

Step 7: Consider risk-adjusted value

An AI opportunity can offer substantial upside while introducing significant risk.

That doesn't necessarily mean the organisation should reject it.

It means the business case should account for governance requirements.

A higher-risk AI use case may require:

• additional testing;

• stronger human oversight;

• enhanced security;

• legal review;

• specialist monitoring;

• auditability;

• more expensive implementation.

Those costs should be part of the investment decision.

Governance isn't separate from the business case.

The cost of governing AI appropriately is part of the cost of using AI.

For more on this, read What Is AI Governance? A Practical Guide for Business Leaders 

Step 8: Assign a business owner

Every material AI investment should have someone accountable for the business outcome.

Not just:

Who owns the technology?

But:

Who owns the value?

If an AI initiative is intended to increase Sales conversion, a commercial leader should have accountability for the outcome.

If it is designed to improve Finance productivity, Finance should own the benefit.

If it is intended to reduce Customer Service costs, the relevant operational leader should own the result.

Technology can enable the change.

But business value should have a business owner.

This connects directly with the accountability issues explored in Who Owns AI in Your Business? 

Step 9: Define success before implementation

Before approving the investment, leadership should agree:

What does success look like?

What metrics will we track?

What is the baseline?

When will we measure?

What level of performance justifies scaling?

What result would cause us to change approach?

What result would cause us to stop?

That final question is particularly important.

AI portfolios need stop decisions as well as investment decisions.

An initiative shouldn't continue indefinitely because the organisation has already spent money on it.

Step 10: Build scale gates into the investment

Rather than approving the full potential investment immediately, organisations can use staged investment gates.

For example:

Gate 1 — Problem validation

Is the business problem sufficiently important?

Gate 2 — Opportunity validation

Is AI genuinely capable of improving it?

Gate 3 — Pilot evidence

Can the organisation demonstrate the expected outcome in a controlled environment?

Gate 4 — Adoption evidence

Will employees actually use the solution effectively?

Gate 5 — Value evidence

Is the initiative producing enough measurable value to justify wider investment?

Gate 6 — Scale

Should the organisation now expand deployment?

This changes the investment conversation.

Instead of:

“Should we spend £2 million on this AI programme?”

leadership can ask:

“What evidence do we need before releasing the next stage of investment?”

That reduces risk and improves capital allocation.

The AI Business Case Scorecard

Before approving a significant AI investment, leadership should be able to answer ten questions:

1. What business problem are we solving?

   
   

2. How significant is it today?

   
   

3. Why is AI the appropriate intervention?

   
   

4. What measurable outcome should change?

   
   

5. What is our current baseline?

   
   

6. What is the total cost of ownership?

   
   

7. What adoption level does the business case assume?

   
   

8. How will productivity translate into realised value?

   
   

9. Who owns the business outcome?

   
   

10. What evidence will determine whether we scale, change or stop?

If several answers are unclear, the business case probably isn't ready.

From AI ROI to AI value management

The bigger opportunity is to stop thinking about AI ROI as a one-off calculation.

Organisations need an ongoing AI value management discipline.

That means continually understanding:

What are we investing in?

Why are we investing?

Who owns the outcome?

What did we expect?

What is actually happening?

What value has been realised?

What should receive more investment?

What should be stopped?

This turns AI from a collection of technology projects into an investment portfolio.

And that is ultimately where leadership needs to get to.

Because the most important question isn't:

How much are we spending on AI?

It is:

What business value are we getting in return?

Frequently Asked Questions

How do you build a business case for AI?

Start by defining the business problem, baseline performance and desired outcome. Then assess AI suitability, total implementation cost, expected adoption, risk, measurable benefits and how the investment will translate into realised business value.

How do you calculate ROI on AI?

At its simplest, ROI compares the financial benefit generated by an AI investment with its total cost. However, organisations should include implementation, integration, data, training, governance and ongoing operating costs rather than considering licence costs alone.

What should an AI business case include?

A strong AI business case should include the problem, baseline, desired outcome, expected benefits, total cost, implementation requirements, adoption assumptions, risks, accountable owner, success metrics and criteria for scaling or stopping.

How do you measure AI productivity?

Measure performance before and after implementation using relevant metrics such as time per task, output per employee, processing volume or cost per transaction. Then determine whether productivity improvements actually translate into additional capacity or financial value.

Why do AI projects fail to deliver ROI?

Common causes include weak use-case selection, unrealistic benefits assumptions, low employee adoption, hidden implementation costs, poor baseline data, unclear ownership and failure to translate productivity improvements into measurable business outcomes.

When should an AI project be stopped?

An AI initiative should be reconsidered when evidence shows that achievable value no longer justifies the cost, risk or resources required. Organisations should define these criteria before significant investment.

Is Your AI Investment Creating Measurable Value?

AI activity can grow quickly while leadership visibility of value remains limited.

Stratify's Executive AI Value & ROI Assessment helps leadership teams test whether AI investment is connected to measurable business outcomes.

Use the assessment to examine:

AI investment visibility

Business-case quality

Value ownership

Benefits measurement

Adoption assumptions

ROI visibility

Scale and stop decisions

Because the strongest AI business case doesn't start with what the technology can do.

It starts with the value the business needs to create.

Take the Executive AI Value & ROI Assessment