Agentic AI Is Starting to Pay Off. Here's What That Means for How We Build
For the past few years, the big question about enterprise AI was whether it would ever repay the investment. BCG's 2026 Applied AI Index, based on a survey of more than 1,300 senior leaders, suggests the answer is changing. Almost half of the companies surveyed now report real value from AI. A small leading group, 7.5% of the sample, is growing its top line 2.4 times faster than the bottom half.
The next chapter is agentic AI, meaning software that acts on a goal and doesn't wait for a prompt. We read the research from where we work every day, across web design, UX and AI services. Here's what stood out.
Focus matters more than budget
AI spending has climbed to 3.3% of revenue, and roughly 80% of it now sits outside the IT budget. Marketing, operations and product teams are all buying their own tools. More money doesn't explain who wins. The leaders commit to one multi-year programme, concentrate on a handful of high-value workflows, and almost all of them (95%) track results through KPIs or the P&L. BCG found that companies strong on both strategic clarity and applied AI capability generate five times the value of those that aren't.
Our take. The "chatbot bolted onto the homepage" approach is running out of road. Value comes from redesigning a whole journey, whether that's onboarding, quoting, support or claims. Journey mapping is a design skill, and it should happen before anyone chooses a model.
Agents are where the value is heading, and the controls are lagging
Agentic AI made up 22% of total AI value in this year's sample, up from 17% the year before, and BCG projects 39% by 2030. About 42% of companies expect their agents to act without human approval by then. Today only 5% have the full set of controls in place. Companies with all six controls generate around three times the agentic value of those with one.
Our take. That controls gap is a UX problem as much as an engineering one. Someone has to design the approval step, the audit trail, the "why did the agent do that?" view and the handoff to a human when confidence drops. As agents take on more work, trust becomes something you design into the interface.
Most of the work is people and process
BCG's rule of thumb is 10-20-70. Ten percent of the effort goes into algorithms, 20% into technology and data, and 70% into people, organisation and processes. The research backs this up. Over half of future-built companies (55%) do deliberate workforce planning around which roles should exist alongside agents, compared with 17% of laggards. Nearly 90% of respondents expect AI to create new work, while only about one in ten expects it mainly to replace existing jobs.
Our take. A tool nobody adopts delivers nothing. The teams we see succeed treat adoption as a design problem. They research with the people doing the work, prototype the new workflow and test it before rolling it out. The interface is often the difference between an agent that gets used and one that gets abandoned.
Agent-ready means data-ready
On the technology side, BCG points to the platform as the real dividing line. Among the leaders, 95% are mid-way through a data transformation, and more than two-thirds are committing to a single enterprise-wide AI platform with a common architecture and control layer, rather than a single-vendor stack.
Our take. This part is our own extrapolation from BCG's findings. If agents will increasingly act on behalf of customers and staff, your content, APIs and design systems need to be structured clearly enough for software to use them, as well as people. Well-organised content and consistent components are what make your product usable by agents.
Three questions to ask before your next AI project
- Which end-to-end journey are we redesigning? Choose one that matters to the business and commit to it.
- What happens when the agent is wrong? Define the guardrails, the oversight and the human fallback before launch.
- How will we know it worked? Tie it to a number the business already cares about.
The bottom line
BCG's headline is encouraging. Sector, legacy and starting point don't decide who benefits from AI. Leadership choices do. We'd add that those choices increasingly come down to design. They shape how journeys are built, how people and agents work together, and how much users and staff can trust what they see.
If you're weighing where agentic AI fits in your digital roadmap, we'd be glad to talk it through.
Based on BCG's "The Formula for Agentic AI Value" (Applied AI Index 2026), September 2026. Read the original.
