Measurement made the next question visible | Pensero Blog [Book a demo](../book-demo) [Login](/auth/login/) [Book a demo](../book-demo) [Book a demo](../book-demo) [Login](/auth/login/) [Blog](../blog) / Article # Measurement made the next question visible What working closely with engineering organizations is teaching us about the next layer of AI transformation. ![](/framerusercontent/images/M6IdnKHV4W8TNavnJjAZhuhQvv4.png?width=1600&height=900) ![](/framerusercontent/images/wGNjJxjzpyAbaOsFHtd8JN1eE.png?width=400&height=400) Ivan Peralta · Engineering Oct 13, 2026 ## The Capital We Take From Our Jobs I read quite a lot. Partly because I enjoy it, but also because every now and then a book gives me a different lens to look at something I am already experiencing. This summer, while reading Cal Newport’s *So Good They Can’t Ignore You*, one idea stayed with me more than the others: **career capital**. The idea connected with how I already tend to think about work. A good job is not only about what you do, but about the environment around it: the mission you contribute to, the autonomy you have, the people you work with, and whether you feel you are learning and making progress. Newport added a question I had not been asking explicitly: ![](/framerusercontent/images/XzfxeJCfwzRULxyNtg2MgZYytNE.png?width=3168&height=1344) What will I understand in two years that I do not understand today? What capabilities will I have developed? What kind of problems will I have learned to solve? What perspective will I have gained that would have been difficult to build somewhere else? That question has influenced some of the career decisions I have made over the years. When I joined Pensero, part of the attraction was precisely the position we were in. AI was starting to change how software was built, but nobody really knew where that transformation would lead. Being close to many engineering organizations going through it felt like an unusually interesting place from which to learn. At the time, I mostly thought about that as personal professional capital. What has become increasingly clear to me is that the same thing may be happening at company level. **Pensero is also accumulating capital.** And perhaps the most interesting part is not only what we are building in the product, but what our position allows us to learn about how engineering organizations are actually changing. ## A Privileged Observation Point Pensero puts us in an unusual position. We work with engineering organizations at very different stages of AI adoption, with different structures, cultures and constraints. We see teams experimenting with similar tools and getting very different outcomes. We see what accelerates change, what creates friction, what looks promising at first and what actually holds over time. That gives us something difficult to reproduce from inside a single company: **a cross-company view of how the transformation is actually unfolding.** The product gives us the data and the observation point. Working closely with customers gives us context. Together, they let us move beyond isolated examples and start recognizing patterns. And as our ability to observe the transformation has improved, something interesting has happened: the questions have started changing. ## From “Is AI Working?” to “What Should I Change?” For a while, one of the main questions we heard was relatively simple: **is AI actually improving engineering performance?** Answering that requires measurement. Not adoption numbers alone, but understanding what is being delivered, how it is being produced, at what cost, and with what impact on quality and speed. As the product evolved, we started being able to see much more of that reality: how much work is AI-assisted or agentic, which models teams use, what they cost, where agents operate autonomously, how their work is reviewed, and why different teams may get very different outcomes from the same technology. And once leaders can see that, the conversation naturally moves somewhere else: > **“OK, I can see what is happening. What should I change?”** That has been an important realization for us. Measurement was never the end of the problem. **It was the prerequisite that made the next layer of the transformation visible.** ## The Dashboard Is Starting to Describe an Operating Model Some of the questions we are now helping leaders answer are no longer really metric questions. Which model should we use for which kind of work? When should an agent operate autonomously, and when should a human direct or review it? How much AI budget makes sense for a team? How should we recognize engineers who are particularly good at orchestrating agents? Consider model usage. An organization may discover that a significant part of its premium-model consumption is going into small, low-complexity changes, while some of its hardest work is being done with whatever model happened to be the default. Nobody necessarily designed that operating model. It simply emerged from individual choices. The data can expose it, but fixing it requires a decision about who should use which model, for what kind of work, and with what level of autonomy and budget. The same happens with quality. If agents are producing more code but a meaningful part of that code is later rewritten, the interesting question is not only whether the metric went up or down. Maybe tasks need to be better scoped. Maybe review practices need to change. Maybe a human needs to stay closer to the agent in certain kinds of work. Maybe a different model is a better fit. **These are decisions about how the organization should operate.** That is where measurement starts becoming something more: a way to understand the emerging operating model of an AI-native engineering organization. ## Learning at the Edge of the Transformation One of the things we see repeatedly is that similar tools can produce very different outcomes. Two teams may use the same coding assistant and the same model. One gets meaningful acceleration while maintaining quality. The other produces more output but also more rework. The difference is rarely the tool alone. It depends on how work is structured, how people direct and review agents, the quality of the context they provide, the level of autonomy they allow, and the judgment surrounding the technology. **There is no mature playbook for this yet — and whatever playbook emerges is evolving almost as quickly as the technology itself.** That is what makes the cross-company perspective particularly interesting. We are not trying to predict what the final AI-native engineering organization will look like. I am not sure anybody can do that today. But we can start understanding which patterns are emerging, under which conditions they seem to work, and where they break. This starts to look a lot like the idea of the *adjacent possible* — the opportunities that only become visible once you are close enough to the edge of a changing field. I think helping organizations act on what they measure is one of them. ## From Understanding the Transformation to Helping Run It This is also where conversations with customers are starting to change. Once you can see what is happening, the next challenge is deciding what to do differently — and then making that change work inside the organization. That may mean changing how teams use models, where agents can operate autonomously, how review works, how budgets are allocated, or how engineers develop the skills required to direct increasingly capable agents. And increasingly, customers are asking us to work alongside them through that process — not simply to explain what the data says, but to help turn it into action. In practice, it usually starts with something concrete. Maybe a team starts routing simpler work to a cheaper model. Maybe agents are allowed to open scoped pull requests autonomously in one part of the codebase. Maybe review practices change because agent-generated work creates a different risk profile. Then we look at what happened: did quality improve? Did cost move? Did people actually adopt the new practice? Did the intervention create another problem somewhere else? Some changes hold. Some do not. **Both teach us something.** > Measure. Understand. Redesign. Deploy. Measure again. That is the role customers are increasingly asking us to play: not only helping them understand the transformation, but helping them operate through it. ## When Learning Compounds The interesting part is what happens when this repeats across organizations. Each transformation teaches us something. Some of that stays as experience. Some becomes a reusable pattern. Some becomes methodology. And some should eventually become part of the product itself — a recommendation, a next step, or a better way of interpreting what we are seeing. We are already starting to see early examples of this. If the product can identify that a model is overpowered for a particular type of work, or understand where an organization sits in its progression from AI-assisted development towards orchestrated or autonomous agents, that is more than reporting. It is accumulated judgment beginning to find its way back into the software. That is how the work starts to compound. **Every intervention should leave us — and ideally the product — smarter than before.** Which brings me back to the idea of capital. What I initially thought about as professional capital is increasingly how I think about Pensero too. The product gives us the position. Working closely with customers gives us real-world evidence. And every organization we work with gives us another perspective from which to understand which patterns repeat, which depend on context, and which assumptions do not survive contact with reality. That creates something a single organization would find difficult to build on its own: **an understanding that compounds across many different transformations.** We still have a lot to learn. The operating model is changing while we are trying to understand it, and I expect many of the things we believe today will look different a year from now. That is probably part of the opportunity. If we are intentional about turning those individual lessons into patterns, and those patterns back into the product, **that accumulated intelligence may become one of the most valuable things we are building.** The operating model is changing while we are trying to understand it, and I expect many of the things we believe today will look different a year from now. That is probably part of the opportunity. Total delivery Points delivered 3.3kpts 10% The AI-era engineering performance platform Pensero gives leaders objective visibility into delivery, quality, and AI impact across the organization. [Let's talk](../book-demo) ## Get months of engineering performance data now Stop deciding on gut feel. Get 90 days of objective data in minutes. [Let's talk](../book-demo) Get months of engineering performance data now Stop deciding on gut feel. Get 90 days of objective data in minutes. [Let's talk](../book-demo) Get months of engineering performance data now Stop deciding on gut feel. Get 90 days of objective data in minutes. [Let's talk](../book-demo) [![](/framerusercontent/images/1v1teeWpH0SzUYk5hDKcYFScErY.png?width=180&height=180)](../) © 2026 Platform [Agentic ROI](../platform/agentic-roi) [AI Deployment](../platform/ai-intelligence) [Delivery Intelligence](../platform/delivery-intelligence) new use cases [CEOs](../use-cases/roles-ceo) [CTOs](../use-cases/roles-cto) Company [Careers](../careers) [Blog](../blog) [LinkedIn](https://www.linkedin.com/company/penseroai/) [Support](../support) OFFICES [Barcelona](https://maps.app.goo.gl/6pfxKiKFNKzLPPnB6) [Madrid](https://maps.app.goo.gl/jEpYBUR4E6AqZs4R6) [New York](https://maps.app.goo.gl/szk58Jf2MBq14MgAA) San Francisco SOON legal [Privacy policy](../privacy-policy) [Cookie policy](../cookie-policy) [Terms of service](../terms) [DPA](../dpa) [Security](https://pensero.trust.site/?ph_distinct_id=undefined&ph_session_id=undefined&ph_source=framer_landing) ![](/framerusercontent/images/iXlw4NDLGJLJbTHbLklPOeLqP5o.svg?width=102&height=20) [![](/framerusercontent/images/1v1teeWpH0SzUYk5hDKcYFScErY.png?width=180&height=180)](../) © 2026 Platform [Agentic ROI](../platform/agentic-roi) [AI Deployment](../platform/ai-intelligence) [Delivery Intelligence](../platform/delivery-intelligence) new use cases [CEOs](../use-cases/roles-ceo) [CTOs](../use-cases/roles-cto) Company [Careers](../careers) [Blog](../blog) [LinkedIn](https://www.linkedin.com/company/penseroai/) [Support](../support) OFFICES [Barcelona](https://maps.app.goo.gl/6pfxKiKFNKzLPPnB6) [Madrid](https://maps.app.goo.gl/jEpYBUR4E6AqZs4R6) [New York](https://maps.app.goo.gl/szk58Jf2MBq14MgAA) San Francisco SOON legal [Privacy policy](../privacy-policy) [Cookie policy](../cookie-policy) [Terms of service](../terms) [DPA](../dpa) [Security](https://pensero.trust.site/?ph_distinct_id=undefined&ph_session_id=undefined&ph_source=framer_landing) ![](/framerusercontent/images/iXlw4NDLGJLJbTHbLklPOeLqP5o.svg?width=102&height=20) [![](/framerusercontent/images/1v1teeWpH0SzUYk5hDKcYFScErY.png?width=180&height=180)](../) © 2026 Platform [Agentic ROI](../platform/agentic-roi) [AI Deployment](../platform/ai-intelligence) [Delivery Intelligence](../platform/delivery-intelligence) new use cases [CEOs](../use-cases/roles-ceo) [CTOs](../use-cases/roles-cto) Company [Careers](../careers) [Blog](../blog) [LinkedIn](https://www.linkedin.com/company/penseroai/) [Support](../support) OFFICES [Barcelona](https://maps.app.goo.gl/6pfxKiKFNKzLPPnB6) [Madrid](https://maps.app.goo.gl/jEpYBUR4E6AqZs4R6) [New York](https://maps.app.goo.gl/szk58Jf2MBq14MgAA) San Francisco SOON legal [Privacy policy](../privacy-policy) [Cookie policy](../cookie-policy) [Terms of service](../terms) [DPA](../dpa) [Security](https://pensero.trust.site/?ph_distinct_id=undefined&ph_session_id=undefined&ph_source=framer_landing) ![](/framerusercontent/images/iXlw4NDLGJLJbTHbLklPOeLqP5o.svg?width=102&height=20)