Delivery Management in an Agentic Era

At Crosstide, we recently hosted a live webinar on agentic software development, and a key question that came out was, what does all this actually mean for delivery management? AI is changing how fast teams can produce work - and that shift changes where delivery leaders need to focus their attention. David Titterton was joined by our Technical Director, Steve Thompson, and our Delivery Director, Paul, to dig into the wider topic of agentic engineering and delivery.

We had an engaged group of participants who took part in quick-fire polls and kept the chat buzzing with questions throughout. Here's a look back at the session through my delivery lens, with the key snippets, stats and takeaways.

At a glance: the stats behind the shift

We ran three quick-fire polls during the session:

• 64% of attendees rated their organisation's AI adoption as maturing well, with 29% at a plateau and 7% having started and stalled.
• Asked which stage of the SDLC is toughest to embed AI within, Architecture & Design came out top (64% of respondents), followed by Strategy & Ideation (45%), and a tie between Testing & Quality Assurance and Operations & Maintenance (36% each). As expected, nobody named Engineering itself as the hardest part - exactly the pattern Steve described on the call.
• On measuring impact, 58% told us it's still on their to-do list, 25% have some measures in place, and 17% are already measuring with confidence - a clear, practical next step for most of the group to build on - baselining and defining clear success criteria
  

  In their own words: the challenges and opportunities

  The increased throughput of work from using AI tools, has resulted in a much higher cognitive load on all team members including Delivery Managers. This points back to the age-old challenge of too much work in progress, which has now amplified due to everyone’s concurrent AI chats!

  "Those sort of delivery problems we tried to solve before - limiting work in progress and increasing throughput - I feel like that's shot through the roof now. So the context switching, the cognitive load, I think a lot of these things are contributing to some of that stall, some of that plateau." - Paul, Delivery Director

  Whether AI adoption in your organisation is stalling or not, it’s a good time to remember the agile phrase, “stop starting and start finishing” to keep teams focused on sustainably delivering value.

  "Engineering has been a very... easy, but a very direct way of implementing AI to get certain points of value. But it's everything that's around it - discovery, testing, release - that's the real opportunity now." - Steve Thompson, Technical Director

  A core pillar of Delivery Management is connecting the dots between different aspects of the overall software development ecosystem. The use of agents increases the need for this facilitation and building the connective tissue between the stages, teams and individuals involved in the software development lifecycle.

  My key takeaways 

• Don’t focus solely on the quick wins with AI. Engineering is where AI delivers the quickest wins, but 64% of our poll respondents said Architecture & Design is now the toughest stage to embed AI within (Strategy & Ideation came in at 45%, with Testing & QA and Operations & Maintenance tied at 36%).
  
  Delivery management adds value when orchestrating the stages either side of the engineering process. It does this by translating business and product ideas into right-sized slices of value beforehand, whilst anticipating and managing dependencies both upstream and downstream. The facilitation that Delivery management provides, aims to create a smooth flow of delivery, including any dependencies, handoffs and points of collaboration. This perspective of holistic delivery health is a fundamental success metric to be measured as part of embedding AI further across the SDLC.  
  
• More output, faster = higher cognitive load.  Paul's "work in progress... shot through the roof" line was a standout moment: more features, more reviews, more to ship, all at once. Treating WIP limits and context-switching as a live delivery metric, which can be monitored as agentic engineering multiplies the work being managed by a given team.
  
  It’s clear that Delivery Management needs to monitor, organise and unblock things, to ensure that the work throughput is appropriate for the team to sustain long-term. Optimising and increasing the throughput for a team creates additional capacity. It will be crucial for Delivery Managers to utilise that additional team capacity on strategic outcomes that move the dial for the organisation, as well as keeping a finger on the pulse of how team members are feeling about their workloads, and helping them reprioritise where necessary.

• A shared maturity model keeps everyone pointed the same way. From individual tool use, through team workflows with shared context, to systemic, governed value - knowing, objectively, where a team sits makes the next investment clearer.  A consistent view on the maturity level of teams today, vs what advanced maturity looks like is key in showing a team where they can get to. This also allows key leadership and delivery roles to help progress the team along the maturity curve, by empowering them, creating opportunities and sourcing support and upskilling.

• Measure as you go. 58% of attendees told us measuring AI's impact on their SDLC is still on the to-do list. A simple set of delivery metrics - cycle time, WIP by stage - often shows the balance between cognitive load and a sustainably higher work output.  

  Delivery Leads will generally be monitoring metrics anyway, to understand delivery flow and health. AI can have an impact on the delivery efficiency of a team. However, how much of an impact, whereabouts the impact is and whether the impact is long-lasting, will all be up to the team to decide and factor into their AI implementation. Delivery leads can increase the visibility of these metrics and share the responsibility for improving them with the wider team. 

  A big thank you to everyone who joined us live, asked questions, and took part in our polls - it made for an energising session.

  If you'd like to talk through what AI and agentic delivery means for your software delivery lifecycle, or would like a copy of the full recording, I’d love to continue the conversation - just get in touch with me at anikh.subhan@crosstideconsulting.com.