From the aibl Team
A year ago, you couldn’t find an operator in a regulated sector that would talk publicly about AI. They’re shy at the best of times, and implementations were experimental enough that going on record felt premature. So I wasn't sure what we’d get when we started booking speakers for aiblLIVE.
We got Vikram Lakhotia from Marshmallow on insurance. Ben Lee, Head of AI at Bidwells, on property. Jonathan Behr running automation at Medigold Health in occupational health, and more on the way. These are complex, highly regulated sectors where AI only shows up when it actually works.
That willingness to go on record is what struck me. I've spent time with the sceptical takes on workplace AI, and a lot of them are well-argued. Some of the concerns about risk and misuse are getting more valid as capability increases, not less. But I notice many of the critiques are based on older models or a willful ignorance of their evolution. The technology has moved, and not everyone's assessment has moved with it.
The people speaking at aiblLIVE aren't theorists, and they're not vendors with a customer story bolted on. They've been running AI in environments where getting it wrong matters, and now they're putting their names to what they've learned.
That's what I keep coming back to. A year ago, going public with this work felt like a bet. Now I suspect staying quiet looks like the riskier position.
Early bird tickets are still available. Looking forward to seeing you on April 28th.
Playbook of the Week
The Roleplay Dojo - How to Stop Practising on Your Best Leads
A Deal You Only Get One Shot At
A UK logistics firm lost a £50,000 ARR deal after a polite but unsuccessful meeting. The rep had rehearsed for an hour with his manager, covering objections, the deck, and pricing.
The problem wasn’t the prep; it was the quality of the rehearsal. The manager realised she was too close to the product to play a true sceptic. The rehearsal was polite, but the buyer wasn’t.
The Roleplay Dojo: A Three-Phase Method
The Roleplay Dojo fixes this by separating the character from the critique. First, it builds a realistic Persona of the buyer you’re actually facing. Second, it defines a ruthless Grading Rubric based on what that specific buyer cares about.
The result is a sparring partner that doesn’t just act like the buyer; it judges like them, too.
News worth reading
We’re still collecting stories from the field, what worked, what didn’t, and what surprised you. If you’ve got a case study that belongs on the aiblLIVE stage, get in touch via [email protected].
Two-thirds (67%) of medium-sized UK firms reported a cyber breach last year. The ways in are still basic: unpatched systems, weak authentication, and exposed remote access. Add autonomous agents and the stakes change.
Fraud prevention firm Nametag says 2026 will be the year impersonation attacks take off, with agentic AI pushing the threat forward. The risk isn’t only data theft. It’s action. If an agent is hijacked, it can kick off legitimate workflows like data exports or software deployments, often before anyone notices.
That is the compounding problem. If breach rates stay high, deploying agents doesn’t just automate work. It automates what an attacker can do once inside. The limit on 2026 progress isn’t model capability. It’s whether you can fix access and identity before agents multiply the damage.
According to Techaisle, mid-market firms rank Agentic AI #1 for 2026, but the math doesn't work. The basics lag behind: Data Fabric is #3 and AI-native infrastructure is #5. Agents fail on foundations that get second-tier attention.
The tension is real. Governance of Shadow AI is the #2 challenge. Leaders want speed to market (#9) while battling ROI pressure (#2) and climbing cloud costs. They are trying to reshape work without securing the environment.
What breaks first is the assumption you can skip to Agentic. Treat 2026 as a demo year rather than a build year, and the pilots will stall. Miss the window, and you are left with unscalable toys and dead business cases.
Mid-market leaders are rushing to deploy AI agents while neglecting the infrastructure required to control them. The expert consensus for 2026 is clear: you can’t build an autonomous strategy on broken access control.
With flat budgets projected for 2026, discretionary spending is dead. Funds will flow to AI, but only for initiatives with measurable business cases. This creates a trap: rushing deployment to prove ROI while skipping the hygiene that keeps the organization safe.
Three structural gaps will derail Agentic adoption:
1. Governance: Deploying without policy.
2. Visibility: Planning for a theoretical environment while ignoring actual shadow IT sprawl.
3. Access: Leaving stale credentials active for agents to exploit.
The new year priority isn't buying tools. It is mapping your actual environment, purging old permissions, and aligning AI spend with hard value. Anything else is just automated risk.
Product spotlight of the week
This week we’ve been looking at Prophix One, a finance platform for mid-market teams that have outgrown spreadsheets but don’t want an enterprise suite pulling IT into every change. It brings budgeting, forecasting, consolidation, close and reporting into one finance-led system.
What stands out is its shift from AI as an assistant to AI that does the work. Prophix offers autonomous agents for budgeting and reporting. They can update headcount and OPEX in plain language, then generate tables, charts and variance commentary, with oversight and a traceable audit trail.
That’s where agentic AI will stick first. Finance workflows are well-defined, repeatable, and built for speed without losing control.
Quote of the week
In small companies, AI might be a feature. In big enterprises, AI might be an R and D lab. In mid-market companies, AI can become the engine of the organisation.