9 Mid-Market Principles for Approaching AI

Mid-market firms succeed with AI by treating it as a capability question, not a tool question: start with adoption and culture, not the model. Nine principles guide this: most leaders still underestimate what is possible; cultural buy-in beats technical rollout; intelligence now splits into propositional, procedural, and perspectival layers; bridge-builders inside the business matter more than vendors; go-to-market beats product in B2B AI; move from curiosity to capability by embedding AI into operational workflows; don’t wait for the perfect use case; protect the junior pipeline as senior roles get more exposed; and de-risk every pilot with a pre-mortem. Roughly 78% of AI initiatives that lead with tools rather than enablement fail to reach production, which is why these principles begin with people.

PLUS: An Agentic De-Risking, Pre-Mortem Playbook

From the aibl team

One of our co-founders, Terry O’Dwyer, has been meeting with advisors, investors, agencies, vendors and businesses across the UK. He shared his notes from the field and here’s a curated version of just a few of the things he’s picked up on the road.

1. Most of Us Just Don’t Get It Over a video call an old colleague walked me through a sophisticated analytics tool that distills thousands of customer records into a sleek reporting interface with real time natural language observations and recommendations. What’s shocking is that he’s a career product manager, not a programmer and he built the tool with a mix of agents and AI-crafted code. It was a project that would have had my last CTO projecting 3 months and tens of thousands in budget. It took 30 hours…and it was his first time building anything like it.It can be comforting to find the flaws and limitations in AI, but pretending that it’s not raining doesn’t keep you dry.
2. AI Lacks Perspective, Not Power AI can handle data (with guardrails), scale and repeatability but it’s not adept at determining relevance in context. The true differentiator in the AI era will be human judgement, framing and the ability to decide “what counts.”
3. SMEs Need a Smarter Build-vs-Buy Strategy Csongor Barabasi’s quadrant offers a clear way forward:
   • Build = strategic IP
   • Buy = non-core tools
   • Hybrid = tailored infrastructure
   • Avoid = shiny distractions This helps cut through noise and reduce AI paralysis.
4. The Three Types of Intelligence We’re shifting to a three-layered “intelligence stack:”
   • Propositional (data/facts) = AI
   • Procedural (know-how) = AI + Human
   • Perspectival (why it matters) = Human

   This is reshaping job design and leadership priorities. This reinforces the importance of helping junior employees add value to the ‘why it matters’ layer as quickly as possible. It used to take years…how do we cut that down by accelerating their learning curve?

5. Cultural Buy-in Beats Technical Rollout In past waves of disruption, we expected 75% of digital transformation efforts to fail and technology was rarely at fault. It’s no surprise that this sounds a lot like some of the studies into AI pilot ROI, because the reasons are similar. As Drucker was supposed to have said, but didn’t, ‘culture eats strategy for breakfast.’
6. AI Champions Accelerate Adoption Companies that embrace AI have internal “bridge builders”, employees who translate between tech and teams. Empower these people early and celebrate them. They’re often not in leadership positions and may not be used to building consensus or enrolling teammates across teams…support and teach them how!
7. GTM > Product in B2B AI Investors are aligned: distribution and execution are the new moats. The best product doesn’t win. The best adoption does. Think enablement, not just features. An AI product that can do something unglamorous but do it well every time trumps cutting-edge reasoning.
8. From AI Curiosity to Capability SMEs must go beyond testing AI tools toward embedding AI into operations, decision loops and workflows. Curiosity is the spark. Capability is the fire.
9. Don’t Wait for Perfect Use Cases Perfectionism and the pursuit of the ‘big idea’ are blockers for now. Start small. Use AI to accelerate admin, automate repeatables and empower customer-facing teams. Learn by doing.

Playbook of the week

It’s a caffeine-fueled brainstorming meeting and the ideas are flying. But you’re that person who considers opportunity cost or asks “if this is such a great idea, why aren’t our competitors doing it?” This playbook is for you…the smart skeptic.

The “AI Pre-Mortem” for Risk Evaluation

Goal: To identify and mitigate all potential failure points before launching a new strategic initiative (e.g., a new product, a market expansion).

Data/Inputs: At a minimum, give the AI your 1-page strategic plan or project brief for the new initiative. Better yet, give it your strategic plan and the project brief, as well as the notes from any planning meetings so it can weigh the potential versus the strategic fit.

AI-Driven Playbook (Step-by-Step):

1. Prime the AI: “We sell X to Y markets. You are a member of my strategy team. We are about to launch [Project X]. Your specific role is to be a ‘pessimistic expert.’ You are analytical, risk-averse and brilliant at finding hidden flaws.”
2. State the Scenario: “I want you to perform a ‘pre-mortem.’ Imagine it is 12 months from today and [Project X] has failed. It has been a total financial and reputational disaster.” (We recommend playing with the language here – sometimes it can be more effective to use weaker terms, sometimes not…that’s the fun of black box AI!)
3. The Key Prompt: “Tell me the story of how it failed. What went wrong? What assumptions were incorrect? What external factors did we miss? Be specific.”
4. Synthesize and Prioritise: After the AI generates the failure narrative, ask: “Group these failure points into 5 key themes (e.g., ‘Market Misunderstanding,’ ‘Technical Failure,’ ‘Internal Misalignment’).”
5. Develop Actions: For the #1 theme, ask: “What is the single most effective action we can take today to mitigate this risk?”

Strategic Output: You move from a “best-case scenario” plan to a resilient, “battle-tested” strategy with a pre-built risk mitigation plan.

Optional Booster: Create a document that captures what you’re good/bad at as a company or for the specific teams involved in the initiative. Great at ideation, slow to deliver? Rock solid with the core business but less successful at expanding business lines? How about estimating time and budget for new products? Feed this document up front to help shape the analysis. Just make sure the AI is private or your work is anonymous before getting started.

NEWS

Before we get to this week’s news, a reminder that AiBL Live London ‘26 is launching soon, and we want your stories. We want to hear about your massive wins and learn from the things didn’t go according to plan. We’ve all been there.

Maybe you have a case study that is ready for the spotlight?

Whatever your story, I can’t wait to hear it. Drop a line to John@AiBLmedia.com

1. Mid-market confidence is quietly rebuilding as businesses start to weave AI into everyday operations. In BDO’s latest survey of UK mid-sized firms, 57% of leaders said they feel confident about the next 12 months and 54% are already investing in AI, with a further 44% exploring how it could support their work. The firms moving fastest are seeing the clearest gains. More than half now use AI in routine tax processes and 77% expect to maintain or grow headcount as they rebalance skills for a more digital operating model. The issue is not optimism, but follow-through. Until AI is tied to specific workflows rather than broad intentions, many benefits will remain unrealised. Confidence rising is encouraging, but capability rising is what matters next.
2. Mid-market manufacturers reluctant to embrace digital transformation. A newly published report from Forterro examines why Europe’s industrial mid-market is moving more slowly than they’d like toward a more cloud and AI based future. Along with the expected causes of budget and data security, nearly one-third said that senior management simply wasn’t on board. Education about the potential and requirements for a fluid, AI supportive infrastructure is for everyone, even the folks at the top. Manufacturers can feel insulated from change, but this revolution is coming for every sector. (Registration required)
3. Agentic AI reaches the mid-market finance function 
   Agentic AI is becoming integral to finance operations. Sage’s new Finance Intelligence Agent automates data entry, reconciliations and reporting, reducing period close to two or three days and delivering up to 5× ROI. The new Sage X3 platform, built for mid-market firms, extends these capabilities across connected operations, while the Sage Trust Label adds transparency to how AI decisions are made. For finance leaders, it’s a signal that agentic AI is moving from concept to core, turning finance from a cost centre into a strategic driver of growth.
4. GPT-5.1 brings more adaptable AI into day-to-day work. OpenAI has released GPT-5.1 with two modes that shift how people use AI. Instant focuses on quick, natural responses while Thinking tackles harder tasks with more depth. Users can now set tone and style directly, making interactions feel more tailored and more useful. The update marks a move from generic AI outputs to assistants that work the way teams do. For mid-market firms it means faster answers, better reasoning and smoother workflows without extra complexity.

PRODUCT SPOTLIGHT OF THE WEEK

A key partner showed us what they’re doing with Clay the other day, and their experience is worth hearing.

Clay is a platform that pulls in and enriches contact and company data from multiple sources, then uses AI to personalise outreach at scale. Our partners had built a new AI in the loop workflow where new intent signals, like job changes or funding news, automatically trigger tailored emails and CRM updates.

My favorite example from another company was how they’re using the system to evaluate whether a company has poor support documentation and building their targeting and comms from there.

Everything from data collection to message creation happens in one place, so they can test and launch campaigns fast without juggling multiple tools. Critically, it’s easy enough that they’re actually doing the testing that usually gets thrown overboard in the name of deadlines.

Quote of the week

“The more I use AI, the smarter I get, and the lazier that intelligence feels. The results are brilliant. The process feels counterfeit. That’s the paradox of AI. The acceleration quietly saps our endurance. We’re evolving into people who move faster than ever and can’t remember how we got there.”

Bryan Melmed

Frequently Asked Questions

What are the key principles mid-market firms should follow when adopting AI?

The nine core principles are: recognise that most leaders still underestimate AI’s reach; prioritise cultural buy-in over technical rollout; understand the three intelligence layers (propositional, procedural, perspectival); empower internal bridge-builders; treat go-to-market as the new moat in B2B AI; move from curiosity to embedded capability; start small instead of waiting for the perfect use case; protect the junior talent pipeline; and de-risk pilots with a pre-mortem. Together they shift focus from tools to adoption, which is where most AI value actually lives.

Why do most mid-market AI pilots fail to reach production?

Around 78% of failed AI initiatives stall because they lead with tools instead of enablement. The technology works; the operating model, culture, and accountability do not. Pilots that succeed treat cultural buy-in, workflow redesign, and skills development as primary deliverables, with the model itself as a supporting component. Without this, even a strong proof-of-concept tends to plateau once it leaves the original team.

What is an AI pre-mortem and why is it useful?

A pre-mortem is a structured risk-evaluation exercise run before an AI pilot launches. The team assumes the project has failed and works backward to identify why — surfacing assumptions about data quality, user adoption, governance, model behaviour, and commercial impact. It typically takes 60-90 minutes and exposes the failure modes most pilot plans never document. It is the single highest-leverage de-risking step for agentic AI deployments.

What is the three-layer intelligence stack?

The intelligence stack splits cognitive work into three layers: propositional (data and facts) is handled by AI; procedural (know-how and execution) is shared between AI and humans; and perspectival (judgement about why something matters) remains human. This framing is reshaping job design, leadership priorities, and how mid-market firms develop junior talent — because the perspectival layer is where new hires now need to build value fastest.

For the research behind this, see the operator data behind these nine principles.

How should mid-market leaders think about AI ROI in this context?

ROI in mid-market AI should be tracked across three lenses: efficiency (time and cost saved), effectiveness (revenue, retention, decision quality), and capability (new products or markets unlocked). Around 49% of AI pilots fail to deliver measurable returns, usually because metrics were agreed after deployment rather than before. Treat AI investment with the same governance as any other capital allocation: target metrics defined up front, reported on the same cadence as financial KPIs.