I have a question about enterprise agents. When these are created and are supposed to be part of an agent ecosystem, how was that like with your clients: Were these created by the business departments or by engineers? If engineers do it, they have to do it together with the business department, right? I have been wondering what kind of collaboration in your opinion worked best when creating enterprise agents that perform operational tasks. What team composition worked well and what didn't work well? And who "owns" the enterprise agent at the end? IT or business? I see a target picture in my mind's eye but I wonder how companies do it when IT and business are separate departments
Great question! Today I would look to the same pattern used for data, but this is mostly determined by the maturity of the organization. This is mostly due to the governance experience that data teams have and (again, today) this where I would start (IMO management and governance of agents is the focus today and these groups are best positioned.
Soon, however, I see this changing. I expect that agents will be deployed in business processes and at that point the business team will play a more substantial role and will probably own the agents (IT will be support, although even that I expect to change once business people get comfortable building agents themselves (Claude code or cowork, for example, are evolve to become core business toolsets)
That means if a company currently has a centralized data team, it needs to move much closer to the business during the transition period. Do you see it the same way?
That is how I see it, but I have not yet worked in a company that already deploys enterprise AI agents in production. I only imagine it this way, because already when creating dashboards it brought enormous progress when data experts and business experts work together daily, so it is only logical that this will be amplified with enterprise agents.
I see in my mind's eye an "ignoring" of the org chart and a purpose-driven collaboration, and over time a transformation of the org chart, but what exactly that will look like is in my view still unclear.
I am asking because I only imagine it this way and I am looking for people who have already seen it in practice.
Good observation!! Yes, as AI/Agents become much more business friendly under the “Center of gravity” shifting from IT/data teams to business teams. The practical reason is that the tools are now so good that the IT (and some of the data) expertise required to do stuff is rapidly diminished. This will be a big eye opener for tech staff.
Example: we are working with a financial leader and last year we build a sophisticated demo and it took months and that was for a small slice of capability. Just this week we finished the next version of the demo - full end-to-end in 1 week, one person. And the icing on the cake: we were able to turn it over to the business leader who is now making changes on their own (full disclosure: we do need to help them with GitHub but that was it). I would not have believed it, but *I was that developer* who built it all in 1 week!
The governance gap between coding agents and enterprise agents is real. I've seen this firsthand - my coding agent can make autonomous decisions about code commits because the blast radius is small and reversible. But when it touches anything customer-facing or involves real money, the rules change completely.
One thing I'd add: the biggest bottleneck isn't the governance framework itself. It's the handoff point between autonomous work and human review. Getting that boundary wrong in either direction (too loose or too strict) kills the whole value proposition.
The identity management point especially resonates - knowing which agent did what and why is harder than it sounds when you have multiple agents collaborating on the same task.
Really enjoyed this! I totally agree with it. traceability is the key for enterprise agents and for organizations… Clear audit trails give you that straightforward forensic view so you can actually see exactly what the agent did, why, and what rules or data it used.
No more black boxes or guess of what was done. Great piece!
Thank you for the great feedback. Yes, there is a lot of focus today on coding agents, which are very exciting but still are focussed on making an individual productive. Enterprise agents - those that participate in business processes - are quite different. They need “enterprise-grade” capabilities that let agents find each other and safely collaborate, interact, and (soon) transact.
I have a question about enterprise agents. When these are created and are supposed to be part of an agent ecosystem, how was that like with your clients: Were these created by the business departments or by engineers? If engineers do it, they have to do it together with the business department, right? I have been wondering what kind of collaboration in your opinion worked best when creating enterprise agents that perform operational tasks. What team composition worked well and what didn't work well? And who "owns" the enterprise agent at the end? IT or business? I see a target picture in my mind's eye but I wonder how companies do it when IT and business are separate departments
Great question! Today I would look to the same pattern used for data, but this is mostly determined by the maturity of the organization. This is mostly due to the governance experience that data teams have and (again, today) this where I would start (IMO management and governance of agents is the focus today and these groups are best positioned.
Soon, however, I see this changing. I expect that agents will be deployed in business processes and at that point the business team will play a more substantial role and will probably own the agents (IT will be support, although even that I expect to change once business people get comfortable building agents themselves (Claude code or cowork, for example, are evolve to become core business toolsets)
That means if a company currently has a centralized data team, it needs to move much closer to the business during the transition period. Do you see it the same way?
That is how I see it, but I have not yet worked in a company that already deploys enterprise AI agents in production. I only imagine it this way, because already when creating dashboards it brought enormous progress when data experts and business experts work together daily, so it is only logical that this will be amplified with enterprise agents.
I see in my mind's eye an "ignoring" of the org chart and a purpose-driven collaboration, and over time a transformation of the org chart, but what exactly that will look like is in my view still unclear.
I am asking because I only imagine it this way and I am looking for people who have already seen it in practice.
Good observation!! Yes, as AI/Agents become much more business friendly under the “Center of gravity” shifting from IT/data teams to business teams. The practical reason is that the tools are now so good that the IT (and some of the data) expertise required to do stuff is rapidly diminished. This will be a big eye opener for tech staff.
Example: we are working with a financial leader and last year we build a sophisticated demo and it took months and that was for a small slice of capability. Just this week we finished the next version of the demo - full end-to-end in 1 week, one person. And the icing on the cake: we were able to turn it over to the business leader who is now making changes on their own (full disclosure: we do need to help them with GitHub but that was it). I would not have believed it, but *I was that developer* who built it all in 1 week!
The governance gap between coding agents and enterprise agents is real. I've seen this firsthand - my coding agent can make autonomous decisions about code commits because the blast radius is small and reversible. But when it touches anything customer-facing or involves real money, the rules change completely.
One thing I'd add: the biggest bottleneck isn't the governance framework itself. It's the handoff point between autonomous work and human review. Getting that boundary wrong in either direction (too loose or too strict) kills the whole value proposition.
The identity management point especially resonates - knowing which agent did what and why is harder than it sounds when you have multiple agents collaborating on the same task.
Agreed. And thank you for the positive feedback on the article!
Hi Eric,
Really enjoyed this! I totally agree with it. traceability is the key for enterprise agents and for organizations… Clear audit trails give you that straightforward forensic view so you can actually see exactly what the agent did, why, and what rules or data it used.
No more black boxes or guess of what was done. Great piece!
Thank you for the great feedback. Yes, there is a lot of focus today on coding agents, which are very exciting but still are focussed on making an individual productive. Enterprise agents - those that participate in business processes - are quite different. They need “enterprise-grade” capabilities that let agents find each other and safely collaborate, interact, and (soon) transact.