Barbara Arnst, VP Transformation Telenet
(at the time of writing)
It started as agentic coding. It became organisation design.
The whitepaper zooms in on the software delivery lifecycle. In software-heavy companies it is one of the processes, next to customer interaction, document-heavy support functions and IT and network operations, where AI will deliver the biggest gains in the coming years: either more with the same, or the same with less.
Barbara Arnst sums up the essence herself: “We set out to learn agentic coding, but what we actually learned was about organisation design.” That is the thread through the paper, a live loop between micro and macro: what building with agents teaches a single team, and what the same shift means for the organisation around it.
Two things kept coming back. Technical decisions that turned out to be organisational decisions in disguise. And governance cadences, built for human pace, that start to crack once agents work at a different pace.
Barbara’s question to her peers is the one we pass on: do you treat AI as a tool you roll out, or as a fundamental change in how decisions are made?
Enjoy reading
Whitepaper Agentic Delivery at Scale - Telenet x Nova Ventura
We ran the experiment on ourselves first
Rather than advise clients from the sidelines, we tested agentic delivery on our own work. A part-time team of five built ACE Ventura, our internal AI platform, with two goals: rethink knowledge management now that AI sits inside daily work, and build hands-on competence in agentic coding.
The whitepaper documents the choices that shaped the platform: curated retrieval instead of connecting the AI to everything, access based on assignments, one agent per service offering, and full visibility over model choice and cost. It is equally candid about what went wrong, from merge pressure to feature creep. When building feels cheap, you can build the wrong thing faster.
The ACE Ventura architecture
Independently owned modules sit behind one API gateway and are loaded into the main app at runtime, on a shared platform. This modularity is what let agentic coding scale across the team.
Then we stress-tested it at enterprise scale
Part 4 is co-written with Barbara and takes these questions to Telenet, a Belgian telecom company with around 150 teams in a federated delivery model. Together we are working on three connected tracks: AI-powered software delivery across the lifecycle, an operating model redesigned for agentic delivery, and a knowledge layer that stays trustworthy at scale.
Early learnings from the enterprise track:
- Agentic delivery does not create misalignment. It removes the slack that allowed unclear priorities and ownership to remain invisible.
- Tacit knowledge is invisible to agents. If a standard is not encoded, it does not exist for the system.
Four parts and a practitioner reference
- Product decisions and learnings from building ACE Ventura
- What agentic coding taught us about team setup, merge pressure, memory and quality
- How agentic systems will reshape roles, decision flows and team design
- Four leadership beliefs from Barbara Arnst on alignment, adoption, servant leadership and partnership
- A practitioner reference: seven building blocks for AI transformation
The paper is written for executives, transformation leaders, CTOs, architects and product teams who have to make AI work beyond the pilot.
Read it the way it was designed
The whitepaper is 68 pages, designed for print, with several pages laid out as two-page spreads. For the best reading experience, download the PDF (42 MB) and open it in Adobe Acrobat, Foxit or another full PDF reader. Choose View, Page display, Two page view, and enable Show cover page in two page view.
Would you like printed copies for your team? Contact connect@novaventura.com.
Enjoy reading
Whitepaper Agentic Delivery at Scale - Telenet x Nova Ventura
Our AI & Data related services
At Nova Ventura, we help you move AI into the core of your organisation, where it transforms decision-making, operations, products and strategy.
