Setting agentic AI pilots up for success

Over the past few months, I've had many candid, off-the-record conversations with leaders across different industries in scale-ups and mid-market companies. 

A few clear patterns have emerged.

Whilst they don’t have Enterprise sized budgets but they do have speed and agility. The can get a small group people in a room together, pick a workflow and then ship something quickly. I see teams deploy working agents (particularly in sales and customer service) while their larger counterparts tie themselves in knots in endless governance committees. Speed and being close to the customer and work matter a lot.

Small, cross-functional teams with a clear owner, tight metrics, a bias for action and a focus on evidence over opinion deliver best. Prove it works in one area by week four and then expand from there.

At the smaller firms end of the spectrum, teams are still broadly overwhelmed unless they already have a trusted core of technical people in place. When that group exists, adoption takes off because the business already trusts them to make decisions and fix things when they break.

Investors are starting to apply pressure. If you are PE/VC backed you’re probably aleady getting heat on better funnel analytics, pricing discipline and the pragmatic use of productivity agents. (As an aside there might be a new PE thesis emerging » buy labour-intensive businesses and use agents to scale them without adding headcount at the same rate). 

If you're in that smaller business cohort don’t panic, yet. Very few are as far ahead as they make out they are. But do crack on! 

Pick one measurable pain point (renewals, first-response time, lead follow-up). Give someone great ownership and a 90-day runway. Lock down three metrics. Demand a live demo by week four. Keep the scope tight, make the results visible and then keep the momentum going.

Let’s stop running AI pilots that go nowhere.

Most of us are taking the ‘launch fast, learn fast’ approach to AI/Agentic, but it’s not producing meaningful result. A new framework from Hg and Clay gets to the heart of the matter: 95% of pilots fail because the basics - data, process, and ownership aren’t ready.

This week’s playbook looks at what the 5% are doing differently.*

Why we like this framework: It’s straightforward, thoughtful and includes estimates for time and money. Will those estimates apply to you? Probably not, but they’re a good place to start.

Where to watch out: This framework was developed with SaaS companies in mind and focuses on the data issues they face. But data fragmentation is nearly universal and we think this is an adaptable approach.

Start with clarity, not curiosity.

Build a minimal core, not a tech zoo.

Assign a translator.

Where the model’s been applied, firms saw 20-40% time savings in marketing and sales tasks, higher-quality opportunities, and faster sales cycles all from cleaner data and structured experimentation.

AI adoption is less about experimentation and more about discipline. The firms that succeed treat it like process design, not product testing.

For mid-market and scale-up leaders, the message is simple: you don’t need more tools, you need better foundations and one person who owns them.

Check out the full framework.

If you’re working with the widely available agents, you’ve probably noticed they’re quick to help but slow to learn. That’s because most aren’t working with a real brief.

‘Help with research.’ ‘Write a summary.’ “Generate a list of names.’ The results sound useful but don’t earn your trust.

Treat your agent like a vendor. Their value depends on learning your business and needs.

Before you start, decide three things:

A clear role makes every chat more predictable and every result more useful. For example:

Research agent: Reviews background material you paste in and produces short, sourced summaries.
Strategy agent: Helps structure ideas or outline reports, testing different framings before you write.
Project agent: Keeps track of next steps from past chats and turns them into a short weekly recap.

Try this when you’re setting one up:

Why it matters

The moment you define purpose and feedback, your agent stops winging it and starts working with intent. That’s how small hacks turn into habits that prepare you for real, collaborative systems later on.

Mentions are not endorsements. This is a curated list generated by our tool-review agent.  

Best value: Gumloop (AI automation platform) and ElevenLabs (AI voice generator) offer competitive pricing with easy-to-use features suitable for mid-sized teams.

AI CRM: SuperAGI provides AI-driven CRM enhancements with smart task automation and predictive insights tailored for mid-market sales and marketing teams. Attio is built from the ground up for GTM teams with a flexible data model, AI-powered enrichment, workflow automation and custom objects.

AI content and video: Synthesia offers scalable AI video creation and avatar narration, ideal for marketing teams looking to produce engaging multimedia at speed. Descript is an all-in-one editing suite for podcasts, webinars and product demos. Turn recordings into clips or transcripts instantly and collaborate on edits like a shared doc.

Considerations: Browse AI (no-code web data extraction), Fathom (AI meeting insights) and Runway which offers next-gen AI video editing and text-to-video but requires some familiarity with creative workflows.

Don’t rush into complicated setups or overwhelm your team with too many tools. Begin with a simple, reliable workflow focused on a clear goal and limited integrations. Prove value early, then build from there.

1. Google, NatWest, and UK Government partner to accelerate SME AI adoption: Google Cloud’s new collaboration with NatWest and the Department for Business and Trade launches a nationwide AI Works for Business tour, aimed directly at UK SME leaders. It will provide hands-on training and access to Gemini-powered AI tools projected to boost SME productivity by up to 20%, potentially adding £198 billion in total economic value.

2. AI at work: the productivity revolution: We’ve heard for years about the UK’s productivity challenge. AI offers one of the clearest chances to move that needle.

But the lesson from early adopters is that success comes from augmenting, not replacing. AI should act as a colleague - one that never gets tired, always remembers the handbook, and can handle the repetitive parts of the job so your team can focus on higher-value work.

In smaller UK businesses, I’ve seen AI tools deliver productivity uplifts from 27% to 133% simply by automating scheduling, streamlining inventory, or generating first drafts of documents. These are not expensive, high-risk projects, they’re low-barrier experiments that pay off quickly.

3. A’Shadow AI is happening at your company, probably:The use of unapproved AI tools at work isn’t just a big company problem. This new study from Microsoft shows that 71% of UK employees are using ‘shadow AI’ and only 32% are concerned about data privacy.

Does Microsoft have a vested interest in your team consolidating around a tool like Copilot? No doubt, but this study jives with others.

One of the most interesting implications of the research is that choosing an organisation-wide tool is just step one. Only 28% say that their company isn’t providing an AI option, so rigorous implementation, education and a dose of oversight are required to get everyone on-side.

UK mid-sized businesses have outperformed both larger and smaller companies for the seventh consecutive year in labour productivity, measured by average revenue per employee, according to Grant Thornton UK research.

In 2024, mid-market firms generated 11% more revenue per employee than larger firms and 15% more than the UK average for companies with at least 10 employees. This trend, ongoing since 2018, reflects mid-market agility combined with access to financial and technological resources.

Despite these gains, UK productivity remains below international standards and mid-market leaders report significant headwinds. There has been a sharp dip in profit expectations to the lowest level since late 2021 amid rising costs, shifts in the labour market and economic uncertainty.

Skills shortages are the major barrier, with 79% of mid-market respondents citing lack of necessary skills as impacting productivity. Leadership, remote team management and communication skills are identified as priority areas for upskilling.

AI is widely seen as a key enabler of productivity, with 78% of mid-market firms expecting AI to positively impact their operations in the next 12 months. However, cautious optimism remains due to concerns around regulation, transparency and ethical use.

Our takeaway: The UK mid-market remains the productivity engine and AI may be its most powerful lever yet. But its success will depend on human capability as much as technological investment.

Read the full study here.

AiBL has been experimenting with relay.app to create automations that help with productivity. 

We embedded the meeting briefing generator this week, straight from one of Relay's templates. Just connected it up, set the rules and off it went, checking Google Calendar for external meetings, grabbing past notes from Gmail and Drive, and sending a short pre-meeting briefing email with a LinkedIn link to the person we’re meeting.

They have a marketplace of use-case based templates, including slack bots, social media video promoters and competitor reporting tools.  Lots to explore and test. 

For mid-market teams exploring easy productivity boosts, it's well worth a look.

We’re looking for an advisory board, a select group of business, policy and tech leaders looking to help shape how mid-market firms adopt AI responsibly and profitably.

The board meets three times a year to keep our insights grounded in real business priorities and market needs.

You’ll join leaders from the following companies: Mindstone, UKAI, Business AI Alliance, Make, British Chambers of Commerce, Google, Microsoft and many more.

If you’re leading AI adoption inside a growth or mid-market firm and want to help steer the conversation, reach out to [email protected]