> ## Content Index
> Fetch the complete content index at: https://www.frontierprofile.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Graham Penman, CTO and Co-Founder, jaam automation
- URL: https://www.frontierprofile.com/graham-penman-cto-and-co-founder-jaam-automation/
- Published: 2026-10-17T09:37:00.000Z
- Updated: 2026-10-17T08:37:04.000Z
- Description: Graham Penman discusses M, jaam automation's Context Engine for Agentic Work Management, and why context is king when deploying AI agents across enterprise workflows.
- Author: Ewan MacLeod
- Tags: AI Agents, Process Automation, Enterprise AI, Workflow Management, Context Engineering

[Graham Penman](https://www.linkedin.com/in/graham-penman-40141141/?ref=frontierprofile.com) is CTO and Co-Founder of [jaam automation](https://jaamautomation.com/?ref=frontierprofile.com), where he's responsible for both the company's professional services work in automation and the development of M, their Context Engine for Agentic Work Management. With a career spanning document automation, workflow and process automation across legal and financial services, Graham has spent years observing where traditional workflow approaches break down when reality doesn't follow the diagram. That experience led him and his team to build M from the ground up - not as a platform with agents bolted on, but as a system designed for people, agents, applications and systems to work together across end-to-end business processes. In this interview, Graham explains why context is king, why automating chaos just increases the chaos, and how enterprises should be developing a context strategy for AI today.

My questions are in bold - over to you Graham:

---

**Who are you and what's your background?**

I'm a software engineer by trade and I've spent my whole career in automation.

I started in document automation, helping law firms and financial institutions automate the production of complex documents. From there I moved into workflow and process automation, looking at how you connect individual pieces of automation into wider business processes.

Over the years I've worked as a developer, consultant, in product and in pre-sales, before co-founding jaam automation.

The common thread has always been a fascination with how technology can remove friction from work. I love building software, but particularly software that helps people work more effectively. That journey ultimately led us to build M, our Context Engine for Agentic Work Management.

**What is your job title and what are your general responsibilities?**

I'm CTO and co-founder of jaam automation, so I'm responsible for everything technical across the business.

That covers our professional services work, where we design, build and support automation solutions for customers, as well as our product and engineering team building M.

M actually grew out of the patterns we kept seeing in our customer work. We couldn't find a platform that managed work in quite the way we thought was natural, particularly as AI agents started to change what a workforce could look like.

Today I set much of the technical and product direction for M, while making sure we're keeping pace with what's happening in AI and doing it in a secure and governed way. I'm still a software engineer at heart, though, so I like getting my hands dirty technically as well.

**Can you give us an overview of your business?**

There are two sides to jaam.

Our professional services business helps organisations identify where automation can make a difference and then design, implement and support those solutions. We work particularly heavily in legal and financial services.

Then there's M, the product we've built from everything we've learned doing that.

M is a Context Engine for Agentic Work Management. It lets organisations manage end-to-end work across people, systems and AI agents as one workforce.

The key is context. Every organisation has its own processes, policies, language, knowledge, responsibilities and ways of making decisions. Humans learn that context naturally when they join an organisation. If agents are going to become genuine participants in the workforce, they need access to the right context too.

M provides the platform around that.

**What problem do you solve with AI?**

Most organisations are currently treating agents as point solutions: an agent reviews a document, another researches something, another updates a system.

That's useful, but it isn't the same as running an end-to-end business process.

If an agent becomes part of your workforce, what happens when it finishes its task? What if it doesn't know what to do next? Who does it hand the work to? What if its output needs reviewing? How do you know what it did and why?

That's the problem we're solving with M.

Agents need to become part of the workforce, not another collection of point solutions. That means connecting them into real business processes alongside people and systems, with the right context, policies, permissions, audit and routes for handing work back when something doesn't go as expected.

**What does the AI actually do, and how is it built?**

Supplier onboarding is a good example.

A supplier might need contract review, IT and information-security checks, background checks, governance and final approval. Some tasks happen in parallel; others have to happen sequentially.

M manages that process and allocates work based on the skills and context required. We're not starting by asking whether a human or an AI should do something. We're starting with the work.

An agent might perform an initial contract review or background check. A person might make the consequential decision. M coordinates that within the same process.

Technically, we built M from the ground up on Azure, including our own work-item and runtime engine. M wasn't designed to have agents bolted into existing ways of working. It was designed from the ground up for people, agents, applications and systems to work together across end-to-end business processes.

We use foundation models and agent technology from the Microsoft ecosystem, while M provides the work management, context, orchestration, governance and audit layer.

M has autonomous background agents providing process intelligence, as well as a conversational agent that can help users understand and act on their work. We're also opening that up so customers can build their own agents using M's tools, skills and knowledge, and connect external agents through technologies such as MCP and agent-to-agent connectivity.

**Who are your target customers, and what's your revenue model?**

Financial services and legal are natural markets for us because of our experience there, but M is relevant to any organisation trying to move towards greater autonomy in how work gets done.

We don't think organisations jump from manual work to fully autonomous agents overnight. There's a journey: digitise the work, use AI to understand and assist with it, allow agents to own defined tasks, and gradually increase autonomy where the evidence supports it.

M is sold as SaaS across Team, Business and Enterprise plans. Rather than focusing primarily on per-user licensing, plans include consumption credits for running work and using the AI capabilities within the platform.

We've only just launched M, so we're early in that commercial journey, with existing jaam customers already exploring adoption.

**What's the evidence it works, and what have you learned along the way?**

We've spent years implementing process automation for real customers, so we've seen where it works and where it breaks.

One of the biggest lessons is that process is good until you hit your first edge case.

Real work is rarely linear. Things happen in parallel, get sent backwards, require somebody to intervene or encounter an exception. We've repeatedly found that traditional workflow approaches can become very difficult when reality doesn't follow the diagram.

Those patterns are a big part of why we built M.

Because M has only just launched, I'm not going to pretend we have years of product metrics. The evidence today comes from years of solving these problems and the reaction we're seeing when customers recognise their own processes in what M can do.

We've also learned the importance of openness. M has its own user experience, but we've invested heavily in APIs and reusable components so customers can also use M as the underlying work engine for their own applications.

**How do you handle the data question?**

We're built on the Microsoft stack, and security and data segregation have been designed into M from the outset.

We don't train our own models on customer data. We use existing foundation models through our Azure environment, and customer data isn't used by us to train those models.

Tenant segregation is fundamental to the architecture. Agents don't simply get access to everything because they're AI; their access to information and actions is governed by the context in which they're operating.

Policy is becoming increasingly important too. We have policies governing our agents today and we're building our policy engine so customers can layer their own AI policies and governance requirements over agents operating within M.

We're currently working towards ISO 27001 and completing Cyber Essentials.

For us, governance isn't something you add once AI becomes powerful. It needs to be there from the beginning.

**What's new at the company, and what's coming next?**

We've just launched M and have an 18-month to two-year roadmap for where we want to take it.

A big part of that is increasing autonomy.

Today you can define a process in M: its stages, tasks, rules and allocations. M then runs that work across people, systems and agents.

The direction we're heading is to reduce how much needs to be explicitly predefined. Instead, you could describe the process, intended outcome, policies and boundaries, and agents could dynamically determine what tasks need to happen, when and who or what should do them.

That's ultimately where we want to get to: agents running much more of an end-to-end process.

But autonomy doesn't mean removing governance. It means moving from prescribing every step towards governing the outcome and the boundaries within which agents can operate.

**If you had a magic wand, what one thing would you change about how AI gets bought and deployed?**

I'd get organisations to think much more seriously about context.

Context is king.

There's enormous potential when you connect AI to systems, give it the right tools and provide good instructions and organisational context. But the opposite is also true.

If you automate chaos, you just increase the chaos.

I used to slightly snigger at the term "prompt engineer". The more we've built with LLMs, the more I've realised how important it is. Two people can use exactly the same model and get completely different results because of the quality of the instructions and context they provide.

Businesses need to think less about simply buying AI and more about curating the context around the work: what does the agent need to know, what can it access, what tools does it need, what policies apply, and what exactly are we asking it to achieve?

**What's your message for the large enterprises evaluating AI right now?**

Prepare your organisation for AI, even if you're not ready to deploy it everywhere yet.

There's a parallel with big data. Having lots of data didn't automatically make that data useful. Likewise, having lots of organisational information doesn't mean an AI agent has useful context.

Enterprises should be developing a context strategy for AI.

Understand how your organisation actually works. What skills will agents need? What systems will they connect to? How will they authenticate? What are they authorised to see and do? What policies govern them?

You can do that work today.

Get the foundations right and you create a much more credible pathway from human-led work, through AI-assisted work and eventually towards greater autonomy.

**Can you list three people you rate from the AI sector that we should be following on LinkedIn, and why?**

I'm probably going to slightly dodge this one.

I'm not particularly influenced by the big AI personalities on LinkedIn. Most of the people shaping my thinking are the engineers I'm working with and the people inside our customers who are actually trying to make this technology work against real business problems.

That might not be the best marketing answer, but it's the honest one.

I'm much more interested in someone saying, "Here's a process we actually run, here's where it breaks, here's what we've tried and here's what we've learned," than another grand prediction about where AI will be in five years.

**What AI tools do you personally use day to day?**

Cursor is a big one for us. We've infused AI throughout our development process, from investigating bugs and producing documentation through to exploring ideas and building features.

But we've also invested heavily in context: how our architecture works, our development patterns, Azure DevOps, pipelines and infrastructure. That means the AI understands much more about how we actually build software.

We use Lovable for rapid UI prototyping, particularly when we want to test and refine an idea quickly.

And I use ChatGPT extensively personally, increasingly through voice and dictation and with other tools connected to it.

AI means our engineers can spend less time producing repetitive code and more time on architecture, requirements, testing, validation and the genuinely difficult engineering problems.

**What's the best new AI product or service you've seen recently?**

For me, the most exciting development isn't one particular model. It's the ability to connect agents to tools, systems and other agents.

As an engineer, I've always loved joining systems together and getting them talking to one another. Agents, tool calling and protocols such as MCP make that dramatically easier.

We're moving from chatting to an AI and getting an answer towards talking to an agent that can actually interact with the systems around you and get something done.

I'm experimenting with this personally as well, connecting tools and building agents around routine activities.

It might sound less exciting than naming the latest model or tool, but I think that joined-up nature of AI is where it becomes genuinely transformative.

**What trends do you think are going to define AI over the next year or so?**

Context is going to become a huge topic, and I think we're going to see agents become much more connected.

Right now, most organisations have agents in silos: one agent doing something over here and another doing something elsewhere. The next step is agents talking to agents, collaborating across work and participating alongside people as one workforce.

I also don't subscribe to the idea that AI means SaaS is dead.

AI can make it incredibly easy to build an impressive prototype, but there's a huge difference between a prototype and safely running production software. You still need multi-tenancy, data segregation, identity, security, reliability and governance.

I think SaaS evolves rather than disappears. Generic products that don't embrace AI will struggle, but good SaaS will become more focused, more usable and increasingly agentic.

It's one thing to build a prototype. It's another thing to run it safely.

---

Many thanks to [Graham](https://www.linkedin.com/in/graham-penman-40141141/?ref=frontierprofile.com) for taking the time to share his insights with Frontier Profile. You can learn more about [jaam automation](https://jaamautomation.com/?ref=frontierprofile.com) on their website.