From the AiBL team
Last week I mentioned that our next research project is on the cultural barriers to adopting AI, and it’s already gotten interesting. I thought I’d share a few smart quotes from some early interviews.
“The biggest cultural predictor is how curious the leaders are. If a CEO isn’t personally engaging with the tools, everything becomes a slow, bottom-up revolution.”
“AI still isn’t tangible enough for most executives. Without their own experiments, it’s impossible for them to see the potential.”
“One divisional CEO is building agent workflows on weekends, and another isn’t even comfortable with analytics or CRM. That gap drives completely different adoption speeds inside the same company.”
“For a company with a largely junior workforce, the challenge isn’t offering tools—it’s getting people to use them for anything beyond replacing Google search.”
“Vendors are pitching capabilities that are months ahead of reality—because they’re building the product while they’re selling it.”
“Many agent vendors promise ‘no IT involvement,’ but that’s fantasy. The moment you try to implement anything real, IT has to get involved—and it becomes three extra projects for them.”
“People are being asked to adopt new tools without taking anything off their plate. It’s AI on top of the day job—not instead of the day job.”\
We’re excited to capture more insights like these, so if you’re interested in participating, do let me know at [email protected])
Playbook of the week
Setting up a continuous feedback loop to capture performance, cost and user satisfaction from AI
While everyone else is geeking out about new AI tools, you’re the one thinking about how to implement real change. You ask the hard questions like “How do we learn from our experience and improve performance over time”?
You think about operations, and this playbook is for you.
As agents move from novel pilots to integrated tools, we want them to get smarter, safer and cheaper. Like any implementation, this requires a process that generates and monitors user data to mine business insights.
In this playbook, we’ve broken down a framework into three stages.
Setting up a success tracker, or “What should we measure”?
Identifying how to go beyond logging failures to correctly categorising them and knowing which team owns the fix.
Finally, using the loop to inform resource allocation, balancing potential gains against effort.
The inner workings of AI may be opaque, but by using a disciplined approach to its output, we can ensure a safe and powerful implementation.
News worth reading
Before we get to this week’s stories, another reminder that AiBL Live London ’26 is approaching and we’re still collecting real-world wins, misfires, and everything in between. If you’ve got a case study that deserves a spotlight, we want to hear it.
Drop a line to [email protected]
Mid-market firms are now putting 10–12% of IT budgets into security, with good reason
New analysis shows companies with 250–1,000 employees are now spending 10–12% of their IT budgets on security, driven by rising breach costs (averaging around $5m per incident, according to IBM) and a growing list of compliance requirements. Attackers are increasingly targeting the mid-market - “big enough to tempt attackers, too lean to burn cash like a Fortune 500,” according to the report. For teams exploring agentic AI, the takeaway couldn’t be clearer: AI only works when the basics work. Identity, access and logging have become the entry ticket for automation, not an optional extra for later.Citing Salesforce two weeks in a row feels lazy, but their latest update really matters for mid-market teams trying to safely operationalise agents. The new observability features in Agentforce give organisations clear dashboards showing what agents are doing, where they stall, and whether they’re staying within guardrails. For firms that don’t have the resources to build their own monitoring layer, this closes one of the biggest adoption gaps. It’s a practical step toward making agents a trustworthy part of day-to-day operations.
Worldpay has introduced a standard that lets AI agents place orders without human involvement. UK research shows the shift is already underway, with 45 percent of shoppers planning to use AI agents and 58 percent open to letting them complete purchases. The the Model Context Protocol is the first serious attempt to define how those agents behave at checkout.
The bigger shift is that the buyer stops being a person and becomes an agent. Legacy checkout flows and fraud checks rely on human signals like clicks and dwell time, and agents don’t generate the same signals. Autonomous buyers will gravitate to suppliers whose systems are clear and predictable, so mid-market firms that adapt early will be the ones agents recognise and select. (And yes, between Salesforce’s monitoring tools and Worldpay’s protocol, the plumbing layer really is getting the headlines this week.)
Product spotlight of the week
This week we got a look at HiBob and it fits the theme of essential but unglamorous plumbing: an HR platform that helps mid-sized firms move beyond the “post-spreadsheet” phase where payroll apps can’t manage performance reviews and onboarding becomes a mess.
The interesting part is not the tool itself but how they use AI internally. More than 90 percent of staff use ChatGPT Enterprise and have built over 2,500 internal GPTs to run the business. That internal-first approach is now shaping the product, from an AI co-pilot for payslip queries to a Growth Coach for performance feedback and AI-driven insight on attrition and compensation patterns.
Their newest module, Bob Finance, connects HR and finance data so teams can spot trends, explain variances and run simple what-if scenarios. It is a clean example of where the mid-market stack is heading: one intelligent system to handle the real operational knots as firms grow.
Quote of the week
“AI is brilliant at the boring things. It’s not yet making dull work creative, but it’s making creative people less weighed down by the dull work. That’s where its value is today…decluttering admin, speeding up repetition, and clearing the small hurdles that slow everything else.” .
”