FocusNow AI Club, meet-up #1 – 29 April 2026.
We started the FocusNow AI Club with a deliberately unglamorous format: everyone around the table, sharing what they actually do with AI today and what they want to learn next. No slides, no vendor pitch. What came out of it was the most useful thing a consultancy can have – an honest internal read on where AI is already earning its keep inside ServiceNow, and where the marketing has run ahead of the platform.
The SPM question worth asking out loud
The sharpest observation of the session concerned Strategic Portfolio Management. ServiceNow has spent three years putting AI at the centre of its messaging, but that value is not evenly distributed across the platform. In ITSM the case is strong: high ticket volumes, repeatable and atomic tasks, predictable patterns. That is precisely the terrain autonomous agents are good at, and auto-resolution and auto-assignment at scale will change how service desks are staffed.
SPM is a different proposition. Ask what you would confidently deploy to a large enterprise portfolio today and the list gets short quickly. Summarisation – the headline capability – solves a problem most portfolio managers do not have. If a demand or business case matters, they will read it. If it does not, a summary does not change the decision.
We think that gap closes. But we would rather tell an SPM client where the value is thin right now than sell them a roadmap built on a demo. That honesty is the point of running the club in the first place.
What is already working on the platform
The team mapped ServiceNow’s AI capabilities as they have actually encountered them in delivery:
Virtual Agent handles the high-volume, low-complexity end of IT support – password resets, guided request flows – using NLP to interpret the user’s query. Mature and predictable.
Predictive Intelligence auto-assigns tickets to the correct group, categorises incidents and predicts priority. The constraint is data: it needs substantial clean history in the instance before the models are worth anything, which rules out a large share of implementations.
Now Assist generates incident summaries, resolution notes and knowledge articles. Useful where record quality is good – and this is the catch. Now Assist can only summarise what the customer’s agents actually wrote. Where incidents get closed with “fixed”, the AI has nothing to work with. Data hygiene is now an AI enablement question, not just a reporting one.
AI Search produced one of the clearest wins in the room. Migrating an Employee Center portal from Zing to AI Search gave noticeably better relevance, with control over how results are ordered based on the user’s profile. For clients whose portal adoption is stalling, this is a contained, low-risk change with visible impact.
The pattern across all four: AI in ServiceNow currently delivers most reliably at the conversational and search layer, where the user experience is the product.
Building faster – with guardrails
On the delivery side, the shift is already measurable. One colleague described trying to automate our branded document with Claude Code eight months ago and failing – the nested table structure defeated it. The same task, attempted again in the week before the meet-up, took a couple of hours. Interestingly, the hard part was not mapping content into the right sections; it was paragraph formatting and table borders.
That eight-month delta is the real signal. Capabilities that failed a feasibility test last year are worth retesting now, and any client whose AI assessment predates this year is working from stale conclusions.
For platform work specifically, external LLMs can now generate UI actions, business rules and flows from plain-language requirements, and a crop of vendors are offering ServiceNow-specialised models that connect directly to an instance. Two cautions came from people doing this daily. First, requirements discipline matters more, not less – the model needs a clear, well-formed story to work from. Second, these tools confidently invent ServiceNow classes that do not exist. Catching that requires deep platform knowledge, which makes AI a force multiplier for experienced consultants and a genuine risk in the hands of juniors reviewing their own output.
Licensing was raised as the commercial counterweight – App Engine’s per-table model, field limits and annual revisions all shape what is economically sensible to build, regardless of how fast AI makes the building.
Key takeaways
- ITSM and SPM are not at the same maturity level. Deploy agents where volume and repeatability are high; be sceptical of AI-led SPM business cases for now.
- AI Search on Employee Center is a fast, contained win for clients with underperforming portals.
- Now Assist and Predictive Intelligence are data-quality dependent. Assess the instance before promising outcomes.
- Retest anything you ruled out more than six months ago. The capability curve is steep enough that old feasibility findings expire.
- Experience is the safeguard. AI accelerates consultants who know the platform well enough to catch its mistakes.
Join the next one
The FocusNow AI Club meets monthly. Next sessions move from round-table to working demonstrations – real use cases, built and shown, with an open discussion of how implementation roles themselves are changing as analyst, developer and architect boundaries blur.
If you are wrestling with where AI genuinely fits in your ServiceNow roadmap, we would like to hear from you. Get in touch with the FocusNow team.