What Advertising Can Teach Us About Making AI Pay

Something interesting happened to the conversation about AI at Cannes Lions this year. People started talking about money.

After several years dominated by what generative AI could create, the discussion increasingly centred on commercial measures such as return on advertising spend, incrementality, targeting, measurement and productivity. Advertising is beginning to provide a useful example of what happens when AI experimentation develops into commercial implementation.

Most businesses have now experimented with AI in some form. Employees are using copilots and large language models, marketing teams are generating content, developers are writing code with AI assistance, and customer service teams are introducing automation. Alongside this activity, businesses have accumulated pilots, proofs of concept and presentations explaining what AI could eventually make possible.

The next stage requires greater commercial discipline, with every serious AI project starting with a measurable business problem. That might be reducing customer service costs, improving conversion, increasing marketing efficiency, shortening product development time or allowing a team to process significantly more work with the same resources. The commercial objective should determine where AI is applied and how success will be measured.

Advertising provides a useful indication of how this is developing because AI adoption has moved quickly across data, content, targeting and optimisation. The conversation at Cannes increasingly focused on whether these capabilities improve return on advertising spend and whether AI-driven activity produces incremental sales.

The same measures can be applied across a business. If AI reduces the time required to complete a process from four hours to one, there is a measurable productivity benefit. An AI-supported customer journey that increases conversion produces additional revenue, while automation that allows customer service teams to resolve more enquiries without increasing headcount reduces cost-to-serve.

Commercial measures also provide executives with something considerably more useful than an AI adoption percentage. A business can have thousands of employees regularly using AI without producing a meaningful return on its investment, particularly where individual productivity improvements are too small or fragmented to change overall business performance.

The quality of the underlying data, systems and processes will have a significant influence on those returns. AI can perform impressively in a controlled demonstration and produce very different results when introduced into an organisation with inconsistent product data, disconnected systems and processes containing years of accumulated exceptions. Scaling AI therefore requires investment in the foundations that allow it to operate reliably.

Advertising is already demonstrating how this can work. Measurement data can be fed back into models to improve future targeting, planning and creative decisions, allowing each campaign to provide information that improves the next one. Ecommerce, customer service, operations and product development can apply the same principle by connecting execution, measurement and optimisation.

Businesses should also examine the workflows surrounding the technology. Giving employees access to AI tools can improve individual productivity, although larger gains are available when entire processes are redesigned around the capabilities now available. Tasks can be automated or accelerated, while decisions requiring judgement, accountability or detailed customer understanding remain with people.

This requires considerably more management effort than purchasing software licences. Businesses need to identify which processes should change, establish reliable data, define ownership, train teams and agree how the resulting performance will be measured. These decisions determine whether an AI implementation becomes part of normal business operations or remains an interesting experiment.

AI projects also need clear criteria for further investment. A controlled test should establish the baseline, measure the result and provide enough evidence to decide whether the capability deserves to be expanded. Projects that fail to create sufficient value should end, freeing investment and management attention for applications producing stronger returns.

The low cost and speed of AI experimentation make this discipline particularly important. Organisations can now create prototypes and test ideas in days that might previously have required months of development. Without clear prioritisation, that advantage can produce dozens of initiatives competing for the same investment, data, technology resources and management attention.

A mature AI strategy therefore needs a relatively small number of projects with clear commercial objectives, agreed measures and defined routes to scale. Advertising is beginning to show what that looks like as agencies change their capabilities, brands reconsider what they expect from partners and practical implementation becomes a larger part of the AI conversation.

The first phase of generative AI gave businesses an opportunity to understand the technology and establish what it could do. Several years of experimentation have now provided enough evidence to make more informed decisions about where it should be applied.

The next phase should focus on where AI creates measurable value, how quickly that value can be demonstrated and whether successful applications can be scaled across the organisation. Advertising is providing an early example of that transition, with commercial measures increasingly determining which AI applications deserve further investment.