An AI feature can make a demonstration more impressive without making a business case more convincing. For vendors selling into enterprise marketing, the useful question is what changes after implementation: the time required to deliver approved work, the quality of a customer interaction, the effectiveness of a campaign or the resources available for another priority.
The July and September US marketing roundtable summaries expose the distance between adopting AI and demonstrating its value. Some contributions described practical productivity gains. Others described restricted access, uncertain measures and difficulty connecting local improvements to business performance. These are accounts of particular organisations, rather than a representative measure of enterprise adoption. Together, they offer a clear challenge for a vendor proposal: explain how the buyer will recognise a worthwhile result.
For The Leadership Board, the commercial implication is to lead with the buyer’s measurement problem. A marketing team may already have access to AI and still lack an agreed baseline, permission to connect the relevant data or a reliable way to distinguish additional capacity from financial savings. A stronger pitch should make those dependencies visible before promising a return.
What measuring AI ROI means in practice
The September business impact discussion questioned the usefulness of cost reduction as the only measure of AI. Contributions considered time to develop assets, conversion, productivity and effectiveness. Customer service examples brought handle time and customer satisfaction into the conversation. Content production offered another practical setting in which teams could observe a change in output.
July’s discussion had already raised a related problem: functional measures do not automatically explain enterprise return on investment. A faster interaction or a shorter production cycle can be useful, but the buyer still needs to understand its significance for the business. The source records uncertainty about how to make that connection, alongside interest in agreeing goals and metrics earlier.
Vendors should therefore ask what the proposed investment is meant to achieve before choosing a dashboard. If the objective is to release scarce specialist time, measure the work those specialists currently do and the work they could take on. If the objective is to improve conversion, define the conversion and the process through which the tool could influence it. Those are different business cases, even when the underlying technology is similar.
This also changes how a vendor should describe AI ROI in marketing. The phrase should introduce an explicit account of costs and benefits, with assumptions that the buyer can examine. A list of possible benefits is a starting point for discovery. It is insufficient as evidence that the investment has paid back.
Establish the baseline before the demonstration becomes a pilot
September’s discussion explicitly emphasised measuring the position before AI implementation. That is an actionable buying signal. A vendor that helps establish the starting point can make its subsequent claims easier to assess and give the internal sponsor a clearer account of what the pilot is testing.
For a content workflow, that starting point could include the time needed to brief, draft, review, approve and publish an asset. This is a recommended measurement approach, rather than a reported roundtable framework. Its purpose is to avoid counting only the step the technology accelerates. A draft produced quickly may still spend considerable time waiting for subject matter review or approval.
The same discipline applies to customer service. Record how the current process handles the selected type of enquiry, including the quality of the resolution and any follow-up work. A reduction in handling time would then sit beside evidence about whether the customer received a useful answer. The source discussions considered both efficiency and satisfaction; a proposed evaluation should preserve that relationship.
Baseline work need not become an indefinite consultancy exercise. Agree a bounded use case, the information required and the person responsible for confirming it. Document any missing evidence. If reliable historical data is unavailable, state that limitation and build observation into the pilot instead of presenting an estimated starting point as a measured fact.
Keep adoption and outcomes in the same conversation
The September summary includes organisations whose initial measures centred on adoption rather than performance. It also records security and compliance restrictions that limited the tools available to marketing. Those conditions help explain why a feature that looks valuable in a demonstration might take time to become part of everyday work.
Adoption is useful evidence about whether an implementation is reaching its intended users. It does not, on its own, establish commercial impact. A vendor should be able to explain what regular use is expected to change and how that change will be observed. This keeps licence utilisation connected to the reason the buyer purchased the product.
The distinction matters during expansion discussions. A sponsor may report that a team has started using the tool while still needing evidence to justify broader investment. Give that sponsor a way to show which workflows are active, which remain blocked and what outcomes have been recorded. An unresolved permissions issue should be visible as an implementation dependency, rather than disappearing into an average usage figure.
There is also a practical discovery question here: does the buyer need another capability, or help making an existing capability usable? The summaries describe organisations with AI access but limited integration into workflows. Vendors should investigate that position carefully. Training, data access or process design may be the immediate requirement around which a credible engagement can be built.
Explain where the saved time goes
September’s discussion described productivity gains and the reallocation of resources within marketing. It also raised the difficulty of calculating specific returns. This combination deserves care. Work completed faster can create useful capacity without producing an equivalent reduction in expenditure.
A vendor should distinguish time released, expenditure avoided and additional commercial activity enabled. These are proposed categories for evaluating a business case. They prevent a sponsor from having to defend a financial claim that was actually based on a productivity observation. If staff remain employed and use the released time elsewhere, the proposal should explain that redeployment rather than calling every saved hour a cash saving.
Ask the buyer what work is currently delayed or left undone. Would the available capacity support more creative testing, faster responses to commercial requests or more time with customers? The source provides examples of resources being redirected, but it does not establish that every organisation will use capacity in the same way. The destination of the time should be agreed within the individual business case.
The September executive AI discussion also describes a new bottleneck around loading creative assets into production. That detail is commercially revealing: accelerating creation can expose a constraint further along the workflow. Vendors should map the route to a usable output and explain which stages their offer covers. Otherwise, a local productivity gain may leave the overall delivery time largely unchanged.
Connect efficiency to effectiveness without claiming causation too quickly
The roundtable material includes accounts of greater creative output and improved engagement. It also includes uncertainty about measurement. These experiences can motivate a useful test, but they should not become a guaranteed relationship between more output and better results. The documents do not supply a controlled basis for such a promise.
A recommended pilot should specify both the operational change and the outcome it is intended to support. For example, faster creative production might allow a team to test more relevant alternatives. The evaluation should then examine the quality of the alternatives, the review effort required and the campaign result. Counting the additional assets alone would answer only part of the buyer’s question.
Vendors should also record other changes that could affect interpretation. A different audience, offer or campaign period may complicate a comparison. This does not make the pilot useless. It means the result should be presented with its conditions, so the sponsor can decide whether it supports expansion, further testing or a change of approach.
That degree of precision can improve the sales conversation. It gives the buyer a result that can survive internal scrutiny and makes clear what remains unknown. An honest account of the limits of a pilot is more useful for an investment decision than a large headline benefit whose method cannot be explained.
Make the cost of implementation visible
The July discussion raised the challenge of making an AI customer service business case when alternative delivery costs were comparable. September added examples of governance restrictions, data silos and unclear use cases. Taken together, these accounts suggest that an enterprise AI business case should include the effort of making the capability work in its intended environment.
For vendors, the recommendation is to set out the expected implementation work alongside the licence or service fee. Identify the customer responsibilities for data preparation, access, review and staff participation. Separate included support from work that would require an additional engagement. The summaries do not provide standard costs, so estimates should be specific to the proposed deployment.
Governance belongs in this account because it can affect both access and the work needed to approve outputs. A tool that requires human review should make that requirement clear. A workflow that depends on information the organisation cannot share should be reconsidered before the pilot starts. Those questions are part of establishing feasibility.
The same approach helps avoid an overextended scope. A narrowly defined workflow may offer a better first test than an attempt to transform the whole marketing function. July’s executive discussion favoured identifying particular bottlenecks. Vendors can translate that into a proposal that states what will be evaluated now and what a later decision would require.
Build a business case the sponsor can explain
The summaries repeatedly connect measurement to stakeholder understanding. July’s discussion included the need to explain technical foundations and to connect AI initiatives with business strategy. September’s discussion considered how marketing budgets might move as AI changed the work. A sponsor needs language that connects those operational details with an investment decision.
A useful proposal should answer a short sequence of questions. What work is constrained today? What will the vendor change? What must the buyer contribute? Which measures will show progress? When will the evidence be reviewed? These questions are a recommended structure for the proposal, not a claim that the roundtables adopted a common procurement process.
Include the decision that follows the review. The sponsor should know what evidence would support broader use and what would justify stopping or redesigning the implementation. Agreeing that in advance gives the evaluation a purpose beyond producing a positive case study. It also helps the buyer allocate the people needed to collect evidence.
The executive summary should be understandable without a product specialist present. Explain the workflow in the buyer’s terms and retain the assumptions behind any benefit estimate. A finance colleague should be able to see whether the proposed value is cash, capacity or a possible improvement in commercial performance. A marketing colleague should recognise the actual work being discussed.
Protect the investments that need a longer view
September’s business impact discussion included an argument for preserving brand building alongside activities with more immediate measures. That is a necessary qualification to an article about proving AI ROI. The source does not support treating every difficult-to-measure activity as dispensable or moving all investment towards the shortest observable conversion.
Vendors should ask how the proposed capability fits the buyer’s broader marketing responsibilities. A tool might improve the speed or quality of work that supports a longer-term objective. The business case can identify those operational improvements without pretending to measure the entire brand effect during a short pilot. Set a review period appropriate to the question being tested.
This is particularly relevant when presenting resource reallocation. If AI releases time or budget, the buyer may choose to invest it in work whose impact emerges later. That decision should be recorded as an intended use of the benefit. It should not be hidden because it does not fit an immediate savings narrative.
What vendors should change in their next proposal
Start with one buyer problem and the current way of handling it. Replace broad productivity claims with a proposed baseline and an observable improvement. Show the route from the tool’s output to work the organisation can actually use. Make the buyer’s implementation responsibilities explicit, and include quality measures wherever faster delivery could create additional review or customer friction.
Then ask whether the proposal gives the sponsor enough evidence to explain the investment internally. The July and September discussions show interest in AI alongside uncertainty about its practical value. The vendor opportunity is to make that uncertainty manageable through a clear use case, a credible evaluation and an honest account of the result.
Related reading
Across The Leadership Board: Bad data only gets fixed when the business can see what it is costing (IT).
Speak to The Leadership Board
Selling AI or marketing technology into enterprise teams? Speak to The Leadership Board about the commercial questions behind the buying conversation, from measuring AI ROI to the implementation constraints your proposal needs to address. Use buyer intelligence to sharpen the outcomes you promise and the evidence you bring.