AI costs put greater pressure on pilot ROI

Cost leads the AI adoption barriers

Cost is the main barrier mid-market decision-makers give for wider AI adoption, according to TXP, ahead of governance and risk, followed by poor data accessibility. That ordering puts the economic case for AI at the front of the channel conversation. TXP also found that just shy of half of mid-market organisations say their AI initiatives have failed to deliver value and have been underwhelming, so customers are judging AI spending by the value it produces as well as the amount they have to spend.

Those findings raise two related questions for partners advising customers. First, a customer has to decide whether it can afford an AI investment. Second, money already committed has to produce enough value to justify wider deployment. TXP reports both cost pressure and disappointing outcomes, but whether one causes the other depends on the individual investment.

Heavy pilot spending meets a 47% progression rate

The value question becomes sharper at the pilot stage because customers are spending substantial sums on AI experiments while TXP reports that only 47% of pilot projects are taken forward by users. A pilot can absorb significant investment and still stop before wider deployment, so progression becomes part of the business case. For a technology leader, the path from experiment to operational use is an economic concern as well as a technical one.

The progression problem leads directly to scaling, where an organisation can establish the basis for AI yet struggle to expand what it has built. Tim Hurst, chief operations officer at TXP, described that distinction directly: “Many organisations have laid solid foundations, but are still struggling to scale AI projects. To be successful, AI needs to know more than the end goal,” His point shifts attention from whether a pilot works in isolation to what the organisation needs to carry it into broader use.

That scaling gap changes how partners should interpret spending. When substantial investment goes into experimentation and many projects stop at the pilot stage, customers need to judge expenditure against what advances into use. Cost remains its own adoption barrier, while pilot progression separately measures whether existing spending provides a basis for further investment.

For channel decision-makers, that distinction creates a practical question before recommending another investment. A pilot can have sound technical foundations and still lack the organisational understanding needed to scale. Finding those missing conditions requires examining the environment around the deployment before treating another round of technology spending as the next step.

Discovery can be the missing investment

That examination begins with discovery, the work of understanding the environment in which an AI initiative must operate. TXP argues that organisations are underinvesting in this phase, connecting the scaling problem to decisions made before and during a pilot. An organisation may know the result it wants from AI while still have an incomplete view of the systems, rules and business requirements that determine whether the deployment can achieve it.

Those requirements depend on concrete organisational knowledge. “Underinvestment in the discovery phase that should underpin AI initiatives means many organisations lack a deep understanding of existing systems, dependencies, internal policies and business goals. Without knowledge of these, AI will never deliver its true value,” Hurst said. Existing systems determine where an initiative has to fit, dependencies identify what its operation relies on, internal policies constrain how it can work, and business goals define the outcome against which value will be judged.

That organisational knowledge also explains why the amount spent on a pilot cannot substitute for adequate preparation. TXP reports substantial pilot expenditure alongside underinvestment in discovery, so customers need to examine where investment goes as well as its total size. A company can commit serious money to AI and still enter a pilot without enough knowledge of the systems and organisational requirements that will determine whether the project works in practice.

Pressure for speed can make that preparation harder. TXP found signs that some firms rushed AI deployments because boards were applying pressure or organisations were responding to the high volume of market noise around the technology. Once speed becomes the immediate decision criterion, teams can reach deployment before they have built the deeper organisational understanding that Hurst associates with successful scaling.

The same need for knowledge continues after the pilot starts because TXP connects organisational understanding with ongoing monitoring. The company warns that project failures will continue unless organisations close knowledge gaps and improve monitoring of pilots as they progress. Monitoring gives customers a way to test whether preparation, operation and the desired business result remain connected while there is still time to adjust the project.

From those operational problems, TXP reaches an optimistic assessment of the barriers themselves. “The good news is that the barriers preventing AI success are practical and solvable, rather than fundamental limitations of the technology,” Hurst said. That assessment directs attention to the conditions surrounding a deployment when a pilot disappoints, including the knowledge available to the team and how the initiative is monitored.

The economic implication is narrower than a claim that discovery will make AI cheaper. TXP identifies weak discovery and insufficient organisational understanding as contributors to project failure, while cost separately leads its ranking of barriers to broader adoption. Poor progression, substantial pilot spending and knowledge gaps can all affect whether a customer demonstrates enough value to support another investment, making discovery relevant to ROI without making it an answer to the price itself.

For channel partners, AI ROI starts before deployment

Because discovery affects the investment case, the channel’s familiar job of demonstrating a clear return on IT investment starts earlier for AI. TXP’s findings point to work during discovery, continued scrutiny through the pilot, and support as the project moves toward wider use. A partner working alongside the customer can reduce deployment risk by understanding the organisation in which the technology will run and carrying that knowledge into later scaling decisions.

That role depends on organisational knowledge because TXP calls for businesses to capture what they know, centralise it and continue building on the resulting knowledge base. For a partner, the work combines technical delivery with an understanding of the customer’s internal systems, information, processes and priorities. That knowledge becomes useful when project choices depend on how systems interact, what information is available, how work is performed and which business outcome the deployment is expected to support.

Hurst makes the resulting partner role broader than technical delivery: “Overcoming them, however, requires more than technical expertise. Businesses must also capture, centralise and build on their knowledge base. A partner that can bring people and technology together, while developing a deep understanding of internal systems, information, processes and priorities, will be central to the next phase of innovation,” His prescription places institutional knowledge alongside technical skill when partners help customers decide how an AI initiative should proceed.

That prescription also matters when judging TXP’s perspective. TXP advocates a larger role for technology partners in discovery, knowledge management and scaling, work from which a partner with those capabilities can commercially benefit. Channel decision-makers can account for that incentive while evaluating the operational mechanism TXP describes: understanding systems, information, processes and priorities before committing further resources to a deployment.

With that incentive visible, applying ROI discipline earlier becomes a concrete approval test. A technically viable pilot needs enough organisational context to show how it can deliver against business priorities and progress beyond experimentation. TXP’s proposed response gives partners a role in carrying knowledge of the customer’s people, processes, systems and priorities from discovery through deployment and into subsequent scaling decisions.

Cost remains a direct constraint

Even with that broader ROI test, the headline barrier remains the price customers face: mid-market decision-makers rank cost first for wider AI adoption. Better discovery can improve the conditions around an investment, while affordability remains a separate constraint that partners have to address directly. The distinction matters because a well-prepared project can still demand more money than a customer is willing or able to commit.

That direct cost pressure shapes the decision partners need to support before more spending is approved. Substantial pilot expenditure has to be considered alongside the likelihood that the work can advance into wider use, and discovery gives the customer information for making that judgment. Organisational knowledge and pilot monitoring can improve the prospects of useful, scalable deployments, while the customer still has to decide whether the expected value justifies the money required.

For the next investment decision, the test is specific to the project in front of the customer. The partner needs to establish what the AI initiative will cost, what business result will justify that expenditure, which systems and dependencies govern delivery, which internal policies constrain it, and how progress will be monitored during the pilot. Those answers give a customer a basis for deciding whether further spending has earned its way into deployment and scaling.

Key takeaways for decision-makers

• Cost leads the AI adoption barriers: Mid-market decision-makers rank cost ahead of governance, risk and data accessibility, while nearly half say AI initiatives have underdelivered. Partners need to connect further AI spending to a clear business result.
• Pilot progression tests AI economics: Only 47% of AI pilots progress, despite substantial spending on experimentation. Technology decision-makers can make progression toward operational use an explicit measure of whether further investment is justified.
• Discovery strengthens the investment case: TXP links underinvestment in discovery with gaps in understanding systems, dependencies, policies and business goals. Customers and partners can establish these conditions early and monitor them throughout the pilot.
• AI ROI starts before deployment: Channel partners can support scaling by building knowledge of customer systems, information, processes and priorities into discovery and delivery. That context gives customers a stronger basis for evaluating whether a viable pilot can produce business value at scale.
• Affordability remains a direct constraint: Better discovery can improve investment decisions, while cost remains the leading barrier to wider adoption. Customers need to assess expected value, scaling prospects and total spending before approving the next stage.