conference
MtPCon London
What stayed with me
Strategy and positioning
- Organisations that aren't using AI wisely across the product lifecycle are going to be left behind.
- Businesses need to get better at strategy: understanding their differentiators and their positioning.
- It helps to codify things: the strategy, what positive impact looks like, what I say no to, and how to make sure customers trust us.
- Businesses need to position for today but have a point of view about the future of the market that centres on their differentiated value, which is where maturity modelling comes in.
Products, interfaces and teams
- When building AI-native products, we need to be clear on who the AI serves, and think about how the customer will view that.
- How products are accessed is changing: SaaS may be dying, and it may make more sense to build products for agents to consume.
- Traditional UI may become a thing of the past; we may need to provide something more dynamic and personalised, possibly chat-based, or something that fits into changing workflows.
- The design role is changing too: designers can build the infrastructure and governance that let agents, and anyone, design, and the QA role is likely to become redundant.
The ideas I kept
Knowing what to build is still the hard part.
Christian Idiodi argued that AI makes building easier but leaves the hardest part, knowing what to build, so the product role should be a conductor of capability, not an author of backlog.
When code is cheaper, bad strategy becomes expensive and bad discovery kills good engineering, which makes discovery the new competitive edge. He asked how much of our work we make happen and how much simply happens to us, and argued for funding stable teams rather than projects, measuring outcomes rather than output, and giving teams problems, not solutions.
Find out early whether there is a research problem.
Jonathan Evens' advice for AI features was to start from pain points AI can solve, redesign the metrics to prove value over the existing experience, and lean on power users to learn as you go. In his view, opportunities are ripest where a task is time-consuming, where AI lowers a barrier, or where it gives people a safe place to try what they can't in the real world.
Find out early whether a research problem is hiding in the idea, start with imperfect but useful evals, and use vibe coding to reach a good-enough MVP.
Jonathan Evens, Building AI products at early stages vs global scale
Specify what trustworthy looks like, or agents will decide for you.
Simonetta Batteiger argued that trust is a product's structural foundation, and that AI algorithms tend to flatten diversity into averages. Agents without clear context default to decisions we never intended, so what trustworthiness and positive impact look like has to be specified clearly, in the product strategy and in the agents' context, rather than left to chance.
The defaults in AI are rarely what you'll want, so design to keep human agency in place and guard against skill atrophy.
Simonetta Batteiger, Don't be an a**hole: how to design AI products that people trust
Ask who the AI serves.
Gary Crawford and Kevin Stewart's Three Futures began from the observation that most dystopias are built from utopian promises, and asked whether an AI serves the human or the system. With every business now a language business, and LLMs working as prediction machines, they expect shifts in what companies sell, how they organise to create value, and how customers reach them.
Customers may stop using our interfaces altogether, arriving instead through a chat interface or sending an agent on their behalf.
Gary Crawford · Kevin Stewart
When agents own the middle, design the edges.
John Moriarty asked what is left of your product when AI handles the job it is meant to do, as people use interfaces less and agentic loops change what products are.
In his view, design shifts from gatekeeper to orchestrator: codify what good looks like in design systems, component libraries and guidelines an agent can use, and redefine done as coherent, intentional and worth the user's attention, because agents amplify whatever direction they're given.
John Moriarty, Building at the edges: when agents own the middle
A product that could threaten yours can now be built in a quarter.
Faith Forster built an enterprise-grade AI product on her own, and reckons it could now be done in three months, which is one quarter in which a new product could threaten yours. Her advice was to layer agents and orchestrate them to create more value, and to use conversational agents to gather understanding before building any interface.
With MCP making data highly transferable, she asked how you make your data more valuable than its source.
Faith Forster, What I learnt building an AI native product
Open the platform to agents, and share the skills.
Axel Sooriah showed how Atlassian's Teamwork Graph, available through MCP or a command-line tool, gives access to every Atlassian product and can be plugged into anything, and how Studio chains skills into agent flows. He also showed skills built into Slack, so the whole team could get insights by calling Cursor with keywords.
Axel Sooriah, Working with AI skills and agents
AI changes the market, not the fundamentals of positioning.
April Dunford said AI does not change the fundamentals of positioning: position against today's competitors but build for tomorrow, with a point of view on the future of your market that centres on your differentiated value.
Positioning evolves as a product moves towards its vision, and each step can be positioned differently, but you can only sell what you have now, so hype the future and connect it to today. When a team disagrees about its point of view, she advises understanding the customer's instead, through a customer advisory board and regular conversations.
What it means for me
- The things I'm good at will continue to be in demand.
- Vibe coding is far more reliable than it used to be, and I need to jump back into it.
- Agents and skills are core competencies, and I need to get more comfortable using them.
About the speakers
Christian Idiodi Speaker
Christian Idiodi is a Partner at Silicon Valley Product Group (SVPG), and has been a product leader for more than 15 years, building teams and developing enterprise and consumer products.
Jonathan Evens Speaker
Jonathan Evens is AI Product Lead at Google DeepMind, and has worked across nuclear fusion platforms, recommender systems and LLM search tools.
Dave Killeen Speaker
Dave Killeen is a product leader at Pendo and the creator of Dex, an open-source personal operating system that works as an AI chief of staff.
Simonetta Batteiger Speaker
Simonetta Batteiger is a product leadership coach and the founder of Inclusive Leaders, with more than a decade of experience leading product and business development teams in technical contexts.
John Moriarty Speaker
John Moriarty is Product Design Director at Fin, formerly known as Intercom, and is based in Dublin. He writes on Substack.
Faith Forster Speaker
Faith Forster is the founder of Discoveree, and a commercially focused product executive and executive coach.
Axel Sooriah Speaker
Axel Sooriah is a Product Management Evangelist at Atlassian, where he helps teams find new ways of working with Jira Product Discovery.
April Dunford Speaker
April Dunford spent the first 25 years of her career as a startup executive, running marketing, product and sales teams, and is the author of Obviously Awesome and Sales Pitch. More at aprildunford.com.