AI adoption is no longer the biggest problem facing many high-tech companies. The harder question is deciding which projects deserve more money—and proving that those investments produce results.

That finding sits at the center of a new Adobe analysis that offers a useful warning for small and midsize businesses: launching a handful of AI experiments is relatively easy. Turning those experiments into a coordinated, repeatable business process is much harder.

Adobe and research firm Incisiv studied 5,633 executives across 12 industries, including 521 in high tech. The research found that 44% of high-tech organizations occupy what Adobe calls the “Frozen Middle.” These businesses have moved beyond testing AI, but they struggle to scale successful use cases from one department to another.

For a small business, that pattern might look familiar. A marketing manager may use AI to draft campaign content, a salesperson may use it to prioritize leads, and a customer service employee may rely on it to answer routine questions. Each effort may save time or improve performance. But if the business uses different data, tools and success measures for each project, the owner may not know which investment creates the greatest value.

Adobe’s findings suggest five practices that can help businesses move from isolated experiments to a more disciplined AI strategy.

Key takeaways

• Assign one person responsibility for AI outcomes across departments.
• Establish a baseline and decide how to measure return before launching a project.
• Give marketing, operations and technology teams shared goals.
• Treat failed experiments as learning opportunities rather than reasons to assign blame.
• Identify employees who can demonstrate useful AI applications to their peers.

The research found that leading organizations were 2.3 times more likely to have a named AI owner with a defined mandate and budget authority. That person does not necessarily need to be a new executive. In a smaller company, the role could belong to the owner, chief marketing officer, operations director or another manager with enough authority to make decisions about budgets, systems and data.

That authority matters when projects cross departmental lines. An automated customer-lifecycle campaign, for example, might require customer relationship management data, product-usage information and access to an email or customer-experience platform. Without a clear decision-maker, a promising project can stall as teams debate ownership or security requirements.

Measurement presents another obstacle. Adobe reported that 44% of its most mature organizations measure return on investment comprehensively, compared with 20% of high-tech organizations overall. The difference often begins before the technology goes live.

A small business testing AI-generated advertising content could record its current production time, cost per asset, conversion rate and revision rate before introducing the tool. It could then compare those figures after a defined trial period. A company deploying an AI lead-scoring system might establish baseline response rates and sales conversions before deciding whether the system improves the pipeline.

Without that baseline, managers may rely on anecdotes. Employees may feel that a tool saves time, but the business may not know whether it increases revenue, improves customer retention or simply shifts work from one team to another.

The report also found a significant gap in cooperation between marketing and technology. Eighty-one percent of the most mature organizations operated at a collaborative or integrated level, compared with 41% of high-tech companies overall. Adobe said 59% of high-tech organizations still relied on siloed, transactional or consultative relationships.

For small businesses, improving cooperation may not require a reorganization. It could begin with a shared objective, such as reducing the time needed to qualify a lead or increasing repeat purchases. Marketing, sales and operations teams can then agree on the data they need, who owns each task and how they will judge the result.

The study identified a further risk in the way businesses respond to failure. Sixteen percent of high-tech organizations displayed what Adobe described as destructive failure patterns marked by blame and repetition. The figure fell below 1% among the most mature organizations.

That distinction has practical implications. An AI-generated product description may contain errors. A customer-segmentation model may target the wrong audience. A chatbot may frustrate customers instead of helping them. Business owners still need safeguards, human review and clear limits on how AI handles sensitive information. But they also need a process for examining what went wrong so the next project does not repeat the same mistake.

The final practice involves internal champions. Forty-two percent of leading organizations said finding and empowering internal champions was their most effective cultural action, ahead of executive sponsorship at 16% and formal training at 11%.

That approach can work particularly well in a small business. An employee who uses AI to prepare better sales briefs or speed up social media production can show colleagues exactly how the process works. Peer examples often feel more relevant than a broad technology mandate, especially when staff members worry that AI tools will add complexity or threaten their roles.

There are limits to applying Adobe’s findings directly to every small business. The research focuses heavily on high-tech organizations, which may have more technology resources and more complex data environments than local retailers, professional-service firms or independent manufacturers. Small companies may also lack the budget to appoint a dedicated AI leader or invest in enterprise platforms.

Still, the underlying management issues apply widely. Businesses need to consider data privacy, customer consent, cybersecurity, vendor costs and the accuracy of AI-generated output before expanding a pilot. They also need to avoid adopting a tool simply because competitors use it.

The central lesson is straightforward: a collection of useful AI tools does not automatically create an AI strategy. Owners should begin with a business problem, define the expected result, confirm that the necessary data is available and decide who will be accountable. They can then test the project on a manageable scale and document what the business learns.

Adobe’s original analysis on AI readiness in high tech provides more detail on how mature organizations structure those efforts. For smaller companies, the immediate opportunity may not be to launch more pilots, but to connect the experiments already underway and determine which ones deserve to grow.

Image via Google Gemini

This article, “44% of High-Tech Firms Are Stuck Scaling AI Beyond Pilots” was first published on Small Business Trends