Jill Sweeney
Jill Sweeney

Businesses are increasing their AI budgets while still working out how to turn adoption into results. A September 2026 Gartner survey found that just 22% of organizations had scaled AI across multiple business units or adopted an AI-first approach. Meanwhile, 85% of functional leaders planned to increase spending that year.

The findings covered 1,303 respondents from organizations with at least $50 million in annual revenue. They suggest a gap between investment plans and operational progress, with Gartner emphasizing the importance of connecting spending to measurable business outcomes.

Smaller businesses face a related challenge. In research published in June, Upwork reported that 74% of surveyed leaders at companies with 10–99 employees said AI had improved productivity. For most, however, those improvements remained below 25%. The findings came from 195 U.S. business leaders.

These surveys measure different populations and cannot explain every implementation difficulty. Together, however, they raise a practical question for buyers. What needs to change in the work itself before another investment can deliver a meaningful improvement?

For Jill Sweeney, chief strategist at Redboard Advisors, a technology and business strategy consultancy, that question should guide purchasing decisions. Drawing on her experience in high-performance computing and AI strategy, she encourages leaders to examine how work gets done, where delays occur, and which processes may need improvement. "If we cannot clearly map out how a process works, we have more work to do before introducing automation," she says.

In her view, buying software can feel easier than addressing unclear responsibilities or discussing the workarounds employees use to keep things moving. When technology teams manage a tool and business teams manage the work, she believes accountability for the overall result can become unclear.

Her suggested starting point is a recurring task with a visible cost, such as processing invoices or preparing customer quotes. She recommends following one real case from beginning to end with the employees handling it, recording delays, repeated checks, and information kept outside the main system.

Before making changes, Sweeney advises documenting completion time, error rates, or cost per transaction. That baseline can help a business judge whether an intervention improves performance. She also recommends identifying one person responsible for the outcome across departments, including the handoffs where work may stall.

That examination may reveal a problem that needs a simpler response, she says. Clearer approval rules, fewer unnecessary steps, or better use of existing software could resolve part of the difficulty. Once those issues are addressed, leaders can assess where AI might help and where human judgment remains necessary.

The risks become concrete in an example Sweeney explains from a software transition she worked on. Deals already in progress were not carried into the new version. For her, the lesson is to test complete business processes and preserve unfinished work, alongside checking that individual software functions operate correctly.

For Sweeney, the same attention to detail extends to costs. Data preparation, integration, computing resources, training and ongoing support all form part of the investment. In her view, forecasts, spending alerts and clear usage limits can help buyers understand how costs may change as activity increases.

For smaller teams, her approach begins with the task that takes up the most time each week. Testing one improvement for 30 days could offer a manageable starting point. She sees value in exploring features already available in existing software, tracking hours saved and having someone review anything customers will see.

"Time saved can become more meaningful when it leads to faster turnaround or gives people more capacity for complex work," she says. "For larger pilots, agreeing on clear measures with the vendor and reviewing progress after 90 days can help both sides decide whether to continue."

She also considers how employees can learn to operate the system themselves. In her view, assigning an internal owner, documenting decisions, and sharing knowledge can help a business maintain the skills needed to manage its work as its relationship with a technology supplier develops.

Ultimately, Sweeney sees readiness in a business's understanding of its workflows, costs, and intended outcomes. That understanding, she believes, can support more informed technology decisions and help leaders recognize when a process change may be enough. One question she uses brings that assessment back to everyday operations. Sweeney says, "Before investing in AI, we need to understand the work, define the outcome, and know how we will measure progress. AI is an amplifier, not a fix. It makes good processes better and bad ones worse "