Using AI as a Work Coach: A Practical Experiment in Leadership Development

I’ve been working with human coaches for years to tackle a challenge familiar to many founders: how do you transition from being the person who does the work to the person who builds the systems for others to do the work? The coaching has been genuinely helpful, but there are practical limitations - cost, availability, and sometimes you just need a quick sounding board for an idea without scheduling a full session.

As our consultancy has grown successfully over the past decade, I’ve found myself navigating the complex challenges that come with scaling - delegation, quality management, and evolving from hands-on delivery to strategic leadership. These are positive problems to have, but they’re still real challenges that need solving.

A recent conversation gave me an idea. What if I could give an AI coach access to all the context a human coach would need - my personality assessment, coaching journey, business challenges, even team feedback - and see what insights emerged?

I’m certainly not the first to explore this approach. Natasha Kenny, an academic leader and certified executive coach, shared her experience of building a AI executive coach late last year, uploading personalized documents including strengths assessments and coaching materials. Companies like Valence are developing enterprise AI coaching platforms, and there’s a growing body of professionals experimenting with AI-assisted leadership development. What I was curious about was how this approach might work for the specific scaling challenges facing a growing consultancy.

The Experiment

I decided to try this with Claude, a family of large language models developed by Anthropic, using a paid plan that includes “project knowledge” functionality - essentially a persistent workspace where I could upload documents and build context over time rather than starting from scratch each conversation.

I fed it comprehensive information about both me and my consultancy:

• Personal context: My CliftonStrengths profile (Relator, Learner, Analytical, Strategic, Achiever), communication preferences, the results of various other personality tests I've done over the years and past coaching session notes.

• Business context: Company performance data, quality management systems, team wellbeing survey results, and strategic documents.

• The challenge: I chose to explore the question of why, despite knowing intellectually what I need to do, I find myself getting pulled away from strategy and business growth and back into operational work.

The key advantage of the project knowledge approach is that each new conversation builds on previous ones. When I have a new issue to discuss, Claude already understands my context, personality, and previous attempts at solutions. It’s like having a coach who never forgets what you talked about last time.

A note on data privacy: I know some people will wonder about sharing sensitive business information with an AI tool. At Research Consulting, we’ve been using Claude to support our client work for some time, so we’d already thoroughly reviewed the privacy protections in place. Claude’s terms indicate that data in paid projects isn’t used for training their models, and you can review their Privacy Center for specifics. Like any business tool, it’s worth understanding the data handling before you start, but having already vetted these protections for client work gave me confidence to proceed with this personal experiment.

What happened next was genuinely interesting.

What Made This Different

Unlike general AI advice or even targeted business consulting, this approach created something closer to a personalised coaching conversation. The AI could connect my individual strengths to specific business challenges in ways that felt both insightful and actionable.

For example, it identified that my Relator strength (building deep, authentic relationships) in the Clifton Strengths framework was likely to create barriers to delegation - because the client relationships I’d built were genuinely personal to me and felt difficult to transfer. This wasn’t a simple failure of delegation technique; it was my natural way of working conflicting with scaling requirements.

Similarly, it connected our team’s wellbeing survey results (which showed excellent work-life balance but highlighted “limited learning time”) to the need to scale our quality management process. The insight: rather than working extra hours to make sure we deliver quality outputs, we need to embed quality as a core part of project delivery from the start - addressing team development needs while building capability.

The Process: Making AI Coaching Work

If you’re curious to try this approach, here’s what I learned about making it effective:

1. Provide Comprehensive Context

Don’t just describe your challenge. Share your personality assessment if you have one, include the results of previous attempts to solve the problem, what’s worked and what hasn’t. The AI can only connect the dots if you give it enough dots to work with.

2. Include Multiple Perspectives

I shared anonymised team feedback, business performance data, and systems documentation, but there are a huge variety of useful sources that could be provided to contextualise a given problem. I found the key was to also include documents and sources that I hadn't written myself to prevent the AI from just reflecting my own biases back at me.

3. Ask for Connections, Not Just Solutions

Instead of “How do I delegate better?”, try “Given my strengths profile and these business challenges, why might delegation be difficult for me specifically?” The insights were far more useful.

4. Iterate and Build

The best insights came through conversation with the AI tool, not single responses. Each exchange refined my own understanding and led to more nuanced suggestions from the AI. This iterative process also prompted me to feed the AI more contextual information - when an insight felt off, I’d add related documents or background to help it make better connections. The project knowledge approach meant this additional context accumulated over time, making subsequent conversations richer and more targeted.

The Broader Opportunity

This experiment convinced me that AI-assisted coaching has genuine potential, particularly for analytical professionals who learn well through structured reflection. It’s not about replacing human coaches - the relationship and accountability aspects remain essential - but about enhancing preparation and deepening self-understanding.

For those of us working in research and scholarly communication, where evidence-based decision-making is second nature, this approach feels like a natural extension of how we already think about problems.

A Personal Reflection

What struck me most was how the AI helped me understand the natural tensions between my strengths and the requirements of scaling a business. That reframe alone was worth the experiment.

If you’re curious about experimenting with this approach - or if you’re wrestling with similar organisational development challenges - I’d be happy to discuss what I learned. At Research Consulting, we work with research organisations on strategic and operational challenges, and increasingly I’m finding that the intersection of individual leadership development and organisational effectiveness is where the most interesting work happens.

In case you're wondering, I also used Claude to help me write this post, asking it to describe the process I had followed in a way that was consistent with my own personality and writing style. It didn't get everything quite right, but gave me a pretty good draft with I then refined the old-fashioned way.