What Realities Remixed Clarified About AI-Native Work
Gijs and I joined Dave Chapman, Esmee van de Giessen, and Rob Kernahan on Capgemini's Realities Remixed podcast to talk about AI-native organizations, developer productivity, vibe coding, enterprise complexity, and what changes when agents become part of how teams work.1
It was my first time doing a podcast.
I expected it to feel more formal. Maybe a bit stiff. The kind of thing where you are too aware of the microphone and spend the first ten minutes trying not to sound like a person trying not to sound nervous.
That did not really happen.
Dave, Esmee, and Rob made it relaxed before we even started. The prep helped. The context up front helped. And once the conversation got going, it stopped feeling like "doing a podcast" and started feeling like a proper discussion about the work we are in every day.
The funny part is that the conversation itself became a good example of the topic.
AI removes some friction.
Then it exposes the next layer.
The question was not whether AI makes work faster
Most AI conversations still start in the same place: productivity.
Can developers ship faster? Can a product manager write specs faster? Can a team generate documentation, release notes, summaries, and reports without spending half the week translating work into updates?
Yes, in many cases.
But that answer is too small.
The more interesting question is what happens after the obvious friction is gone. When writing code gets faster, the bottleneck does not disappear. It moves. It moves into context, ownership, trust, review, governance, and decision-making.
That was the thread we kept returning to in different forms.
AI can help people get into flow. It can reduce interruptions. It can summarize work, update documentation, and turn rough intent into a first version. But if the underlying organization is slow, political, misaligned, or unclear about what it actually wants, AI does not magically fix that.
It just makes the gap visible sooner.
Startups and enterprises feel the same shift very differently
One of the useful parts of the conversation was comparing startup speed with enterprise reality.
Small teams have an advantage because context is still close to the work. Everyone feels the pain. Everyone sees the wins. Decisions can happen quickly because there are fewer layers between the person seeing the problem and the person changing the product.
That closeness is fragile.
As organizations grow, they create systems to protect themselves: process, reporting, approvals, legal review, planning rituals, operating models, steering groups, dashboards. Some of that is necessary. Some of it is just accumulated caution.
The hard part is knowing which is which.
AI makes that distinction more expensive to ignore. A startup can use agents to explore a product idea, ship a prototype, update documentation, and create content around it in a very short loop. A larger company can use the same tools and still get stuck for a month because nobody is sure who owns the decision.
That is not an AI problem.
That is an operating model problem.
Flow state is an organizational issue
We also talked about the ordinary drag inside product and engineering work.
Before One Horizon, a lot of my product management work was information routing. Asking one developer whether something was done. Updating another person that they could continue. Translating progress for a stakeholder. Turning the same reality into a Slack message, a planning note, a meeting update, and eventually a report.
It feels productive because you are busy.
But a lot of it is just moving information between systems that should have understood each other in the first place.
For developers, that drag shows up as interruptions. Slack pings. Status questions. Meetings that exist because the work system cannot answer basic questions. Every interruption has a cost, and the real cost is not the message itself. It is the context switch.
That is one reason we care so much about keeping work connected.
If the roadmap, task, issue, commit, pull request, release note, and recap all live as disconnected fragments, the team needs humans to keep translating. Once agents enter the workflow, that translation problem gets sharper. Agents need context to act. Humans need context to review. Leaders need context to trust what is happening without turning the whole system into surveillance.
The answer is not more status meetings.
It is a better shared record of the work.
AI-native culture is not a tool rollout
"AI-native" is easy to turn into a software shopping list.
That misses the point.
The real shift is cultural and operational. If agents become part of how work gets planned, built, reviewed, and reported, then teams need a different level of clarity around intent. They need cleaner decision trails. They need accountability that does not depend on one person remembering what happened in a meeting three weeks ago.
They also need leaders who understand that AI adoption is not the same as cloud adoption, agile adoption, or digital transformation with a new label.
Those changes mostly altered tools, infrastructure, and process. AI changes who or what can participate in the work.
That makes trust central.
Not abstract trust. Practical trust. Can we see why this task exists? Can we see what changed? Can we see what the agent did? Can we trace the result back to the roadmap? Can a human step in at the right moment with enough context to make a judgment?
If the answer is no, then the organization is not AI-native.
It is just using AI tools.
Vibe coding is not company building
We spent time on vibe coding too, because it is one of the clearer symbols of the moment.
It is genuinely useful that more people can build things now. The distance between idea and first working version has collapsed for a lot of software. That is exciting.
But building a first version is not the same as building a company.
Software does not freeze when the demo works. It needs maintenance, support, security, onboarding, permissions, integrations, documentation, billing, migration paths, incident response, and all the small product decisions that make something usable after the novelty wears off.
That is where a lot of AI excitement gets too thin.
Code generation is part of the story. It is not the whole business. It is not the whole product. It is definitely not the whole organization.
The same goes for product creep. When agents make it easier to build, teams may start adding not just more features, but whole adjacent products. That can be powerful. It can also turn into a mess if nobody is asking whether the work still serves the strategy.
More output is not automatically better execution.
The control layer matters more when creation gets cheaper.
Europe has to answer the speed and sovereignty question
Another thread that stayed with me was the difference between building in Europe and building in the US.
In the conversation, we did not land on a neat answer. The US market often feels more willing to fund the AI scramble. The European market often feels more cautious. Some of that caution is healthy. Some of it makes it harder for European startups to move at the speed this shift demands.
Then there is the sovereignty question.
If company knowledge increasingly flows into agents, where does that knowledge live? Which models are doing the work? Which infrastructure do teams trust? What happens when the operating memory of a company depends on systems built somewhere else?
Those are not abstract policy questions for later.
They matter to product teams now, because AI-native work depends on context. Your company's context is not a side file. It is the thing that makes the work possible.
The most human part of the conversation was the point
There was another layer to the episode that I liked.
It was human.
Not in a polished "thought leadership" way. In a real conversation way. It had jokes, tangents, half-formed thoughts, disagreements, and the kind of small moments that make a discussion feel alive.
That matters more now, not less.
AI can help us create more. It can help us write, summarize, edit, and produce. We use it ourselves. We even built Ink because we wanted a way to make content workflows faster without losing the human context behind them.
But the more content becomes cheap, the more real point of view matters.
That was clear on the podcast. The useful parts were not the perfect sentences. They were the moments where the conversation found a sharper question:
What friction disappears?
What friction moves?
What friction was already there, waiting to be exposed?
What we took from it
The episode was officially about starting an AI company and building in an AI-native world.2
For us, it became a checkpoint.
One Horizon started from a very practical frustration: product and engineering teams waste too much time translating work instead of doing work. AI makes that problem more urgent because agents need the same thing humans need: clear context, clean ownership, and a reliable connection between intent and execution.
That is the layer we are building.
Not another dashboard for the sake of a dashboard. Not a generic project-management app with AI sprinkled on top. A way for software teams to connect roadmap, tasks, issues, commits, pull requests, recaps, handoffs, and release work so people and agents can operate from the same truth.
The podcast made one thing clearer for me.
AI-native work is not about removing every bit of friction.
Some friction is useful. Review is friction. Taste is friction. Judgment is friction. Leadership is friction.
The goal is to remove the wasteful friction: the status chasing, the repeated translation, the hidden context, the meetings that exist because the system cannot explain itself.
Then teams can spend more time on the friction that actually improves the work.
That is the signal I care about.
You can listen to the episode on Capgemini's Realities Remixed page or on Spotify.12 And if the context layer around AI-native software teams is the part you are trying to make sane, that is exactly what we are building at One Horizon.
Footnotes
-
Capgemini. "Realities Remixed - the podcast series." Episode 208, "RR013 Starting your own AI company with Gijs van de Nieuwegiessen and Tijn van Daelen, One Horizon AI." https://www.capgemini.com/insights/research-library/realities-remixed-podcast/ ↩ ↩2
-
Spotify. "Realities Remixed." Episode with Gijs van de Nieuwegiessen and Tijn van Daelen from One Horizon AI. https://open.spotify.com/episode/6f6J4fh4MAP4hSwyzZlLcV ↩ ↩2
