How we're coordinating AI at FELD M

1. What motivated us to write this

Today, the question on everyone's minds is no longer whether AI matters, but rather whether you're doing something useful with it. For most organizations, the answer is: not really.

Having spent considerable time experimenting and building shiny prototypes, we're now coming to a point where the early enthusiasm of the past three or four years is fading away, giving way to a more honest inventory of initiatives.

What organizations are rediscovering, yet again, is that the limiting factor is not technology, but coordination. At FELD M, we've been thinking hard about this. As a data, analytics, and AI consultancy, we feel a responsibility towards our clients: they are asking for our support with navigating complex market environments, helping them to turn data into decisions and, in turn, measurable business value. If AI is reshaping how business value is created — and many of us strongly believe it is — we owe it to you, our clients, to understand that shift and to tackle that opportunity head-on.

This article is an honest account of where we've landed. We're not writing it to announce a finished transformation. We're writing it because we keep seeing the same pattern in our own work and in our clients' work: the shiny wrapping has come off AI, and what's left is the unglamorous, essential work of turning it into something valuable: the architecture, the strategy, the governance that were easy to skip past while the demos were still dazzling.

If that resonates, you're not alone. We'd like to share what we've learned: what worked, what didn't, and how we're approaching the shift from intent to execution. It (hopefully) goes without saying that there's no way to cover this topic for FELD M as a whole, so this article represents the experiences of those of us involved in its publication; we hope you find our perspectives useful.

2. What you can expect from this piece

In this article, you can expect to get some first-hand insights into our path towards AI adoption at FELD M. The article has three parts.

• We start by reflecting on our guiding principles that summarize how we think about AI as an organization. 

• From there, we move to coordination: what we've put in place to help teams make the move from experimentation into production. 

• We close with a collection of perspectives from across FELD M: specialist voices that show where we actually stand, not just where we intend to be. These individual voices serve as evidence for why and how our coordination efforts will bring everything together on an organizational level.  

We hope that some of these thoughts resonate with where you are on your own AI journey. Moreover, we're deliberately sharing our coordination efforts alongside our guiding principles to show you how to deal with current uncertainty.  

3. The challenge: speed versus trust

While the perceived hype leaves many in a state of awe, creating a feeling of fear of missing out, we're deliberately taking a rigorous approach. It's not about speed of innovation alone; it's about focusing on doing the right thing, maybe more than ever.  

In the last weeks and months, we have talked to numerous people in our industry, and we see the same pattern everywhere: speed vs. trust. Experimentation is becoming easy and cheap, but to create sustainable value you need to achieve reliable and repeatable results.  

4. What our responsibility means for you

When we recommend solutions and strategies to our clients, we put our reputation on the line. More importantly, we put yours on the line too: the output of an AI-assisted analysis carries your name inside your organization as much as it carries ours.  

That is why we treat trust as something earned rather than assumed, and why the judgment, the interpretation, and the relationship stay with a human who is accountable for them, something AI will never be able to offer.

Everything we've learned about coordinating AI has come out of real client projects, and we hope you find the perspective insightful for your own internal transformation.

5. Our three guiding principles 

In order to navigate and steer AI transformation at FELD M, we've come up with three heuristics as our governing principles.

1. We adopt AI deliberately, not driven by hype. Focused on what creates measurable value, at a quality we can put our name on.

Quality has always been one of our guiding principles, and we still see this as a key differentiator in a world where content creation and coding are increasingly commodified.

2. AI augments what we do, but the thinking, the judgment, and the relationship with our clients stay human.

3. Every team is responsible for identifying opportunities, but we provide the tools, knowledge, and governance to make it happen.

Our approach to innovation has always been decentralized, yet we understand that structured coordination and enablement are crucial to support our teams. What we provide centrally is the infrastructure: the tools, the knowledge exchange, the governance, and the guardrails to ensure quality, safety, and compliance.

We have decided against a top-down approach that tells teams what to use. Instead, our coordinated approach supports the teams in finding out themselves.

6. Why coordination isn't optional: what our projects have taught us

To establish the right level of coordination to hold things together, we've organized our efforts into two tracks.

• AI services push: Making AI a visible and value-creating part of our client work. This includes AI-assisted analytics and reporting workflows, AI enablement and literacy training, AI governance frameworks and use case methodologies, and helping clients figure out where AI can move the needle in their own organizations.

• AI enablement: Providing the foundation that makes AI adoption possible. This means creating formats for knowledge exchange and learning opportunities, a defined tool stack that supports the use cases in a safe and compliant way, and a governance layer that gives everyone clarity on what's allowed and what's required. 

7. Perspectives on AI at FELD M— as evidence for coordination 

The questions arising from an innovation perspective are manifold. And even in a smaller organization like ours, these are very distinct between the different teams and projects. The idea of our central coordination layer is to make sure every voice is heard and seen; collaboration and co-creation will be a key ingredient to leverage the potential in a meaningful way. 

Here, we have collected some voices to show what these efforts mean across a few key positions at FELD M. 

AI/agentic analytics (Sandy) 

"My experience with agentic analytics really comes down to speed versus trust. These tools can take a plain-language question and turn it into a full dashboard in moments, which takes a lot of manual work off our plate and lets us focus on the business logic that actually matters. But fast isn't the same as reliable. A tool will confidently hand you a number that's completely wrong, and it won't always tell you when a data source is missing or an assumption has slipped in. What I've learned is that it all comes down to the foundation: there's no substitute for a strong semantic layer, clean data underneath, and enough transparency to show its work. The tool worth recommending isn't the fastest one. It's the one you can trust on a normal, messy workday."

Read more about Sandy's experiences with agentic analytics here.

Agentic coding (Joao) 

"I see great advantages in the mental offload while writing and debugging code, the ability to parallelize tasks, and for planning features and performance improvements. What really frustrates me is the unreliability of results, the effort to set up the tools correctly, and the tools wandering off into endless loops, where time and tokens get wasted. Clean guidelines and guardrails, including coding standards, help us improve quality. And for our agentic workflows, it's all about the setup: having them connected to the tools we're using, and having well-defined instructions that are tested for reliability."

Joao will also be speaking on this topic at our upcoming DigiTalk on October 22.

Enablement (Philipp) & collaboration (Sophia) 

"From an organizational perspective, we want to make sure that friction is minimized and teams are enabled sufficiently. This involves regular updates on our guidelines and guardrails to match the rapid development of the market. In addition, we want to empower teams with regular knowledge exchange and learning opportunities."

Business development (Laura) 

"When AI can be everywhere and everything, it is maybe more important than ever to align on what you want it to be. This is especially true when looking at it from a business development perspective: we want to make sure that we are adding it where it creates most value – both for our customers and, as a consequence, for us."

Strategy (Lutz) 

Governance (André) 

"At FELD M, we decided to take a lean approach to AI governance. We have a one-pager that describes what we want to use AI for, how we expect AI to be used from an ethical and technical perspective, and what is explicitly not allowed. We also defined who people can reach out to for specific use cases, questions, issues, or approvals, so everyone knows where to go and who can provide guidance when needed.

Rather untypically, we didn't create fixed processes around this. This is mainly based on two things:

First, we generally don’t have many formal decision-making processes at FELD M. A lot of decisions are made within our collective.

Second, we have established community formats where topics, experiences and concerns can be discussed openly.

So our approach relies on clear ownership, a minimal set of guidelines and the strength of our collective rather than a clearly defined target operating model. This works well for us and the way we operate as an organization. But it probably wouldn't work in the same way for a larger or more hierarchical organization, where clearer processes, responsibilities and decision structures are needed."

Marketing (Eilish) 

"On the operational level, we're noticing great advantages in terms of speeding up routine tasks such as optimizing texts, checking tonality, and spotting basic errors we might have missed using AI.

But going forward, our marketing and content strategy will continue to focus on what AI can't create: Real people, real expertise, real relationships. We started experimenting with naming our specialists more in our content this year, because in a time of increasingly commodified content, showing that real experts are behind our content will become a differentiating factor. We want to outsource the routine stuff to AI, not the creativity.

Our FELD M in Conversation series — full-length interviews with some of the data and analytics industry's most interesting minds — is our answer to a market flooded with generic content. Interviews with real people can't be generated by AI. We're betting that content people relate to on an emotional level, about real people and real stories, will stand out against the rest."

8. Outlook — what comes next 

This is not another success story — not yet. In the current situation, we sometimes feel that we're struggling to keep pace, but one thing we’ve learned throughout our careers is that this feeling of being an imposter is simply part of the game. What matters is not giving in to it, but moving ahead with a clear vision and strategy, while staying open enough to pivot when necessary.  

Our short-term commitments: doubling down on key business opportunities, bringing our tool stack in order, encouraging further learning and collaboration, and updating our governance.

What binds these together is how we are coordinating them. Rather than mandating a fixed playbook or leaving teams to figure it out alone, we've set up a structure that gives clear guidance while leaving room to build and adapt. This lets us move fast without losing direction, especially without losing ourselves and our core principles and values along the way. 

The FELD M way has always been rigorous, pragmatic, and honest - with our clients and with ourselves.

This is how we're approaching AI as well: keeping the human in the driver's seat.