AI in service management is shifting from experimentation to operational accountability. It is no longer sufficient to embed generative capabilities; organizations must ensure those capabilities deliver consistent, real-world outcomes. As vendors expand their AI messaging and product portfolios, buyers should look beyond announcements and focus on a more critical question: what makes AI truly useful, usable, and sustainable in day-to-day service operations? Freshworks’ latest updates provide a timely lens to evaluate that shift.

Context Will Determine Whether AI Service Management Scales — Or Fails

Will McKeon-White, Senior Analyst

Freshworks is the latest provider to expand its AI offensive, adding Agent Studio support for EX use cases and pragmatic improvements: increasing IT AI (including automated change risk scoring), expanding guardrails, automating knowledge generation, and, above all, enhancing the capabilities of its context layer. In addition to the AI injection, improvements to ITOM, ITAM, XLA tracking, and incident management were highlighted. All of these announcements were expected and not necessarily differentiating on their own. But what was different, and what made these AI announcements interesting — the continued emphasis on simplicity and speed of adoption (and the realistic value it generated, validated by customers of all different sizes).

According to customers, big and small, the features were easy to adopt, enabling them to get value from both the platform and the AI. While most adopters were early in their journey, we saw the AI-centric service desk playing out for multiple customers – even smaller shops!  Customers reported that, beyond just deflecting common use cases, it was helping them reshape the role IT plays in their org. Ultimately, it enables AND forces them to be proactive. Instead of being constantly overwhelmed by dozens of basic requests a day, service desk practitioners are helping build better end-user-facing knowledge, identify opportunities to automate, and call out service gaps or chronic pain points. Problem management has long been under-executed due to capacity constraints. AI is now making sustained execution viable.

Organizations pursuing this path cannot afford additional risk — several critical lessons demand immediate focus:
  • Prioritize a tool that your team finds easy to use (and a provider that will help you adopt it). Customers reported that something WILL go wrong, and having both your people able to fix it, and support from a vendor when that gets complicated, is a quick way out of the early implementation quagmire.
  • Don’t expect everything to fix itself right away. AI is a tool to use, not magic. Pick a few pain points in your org to start and align your data strategy around these use cases. Data in this case means process knowledge and system data. Not knowing where data is or how to start is a fast way to sabotage your initiative.  Treat AI as an operational capability, not an autonomous solution — value emerges from targeted use cases, not broad deployment.
  • Start simple and build on success. Pick ten tasks to start with – ten things that cause people pain and are just annoying. And keep building and keep building. Success is additive, not instant.

AI-Ready Platforms Need Two Architectural Bets, Not One

Julie Mohr, Principal Analyst

Freshworks made a deliberate architectural decision when it deployed its own platform internally. Ashwin, the CIO running Freshworks’s customer-zero deployment, mandated no customizations because human-built workflows poison AI consumption. An AI agent trained on workflows designed for human navigation produces a mess at output. The decision was strategic rather than convenient, and it cleared the architectural runway for AI agents to operate against workflows engineered for them. Other ITSM and ESM platforms have not yet made the equivalent bet, which is why their customers spend the first six months of an AI deployment retrofitting around customizations the platform absorbed over a decade of human-led configuration. The vendors that ship retrofitted AI atop human-shaped workflows are selling the appearance of AI readiness, not the substance.

The same architectural discipline has not yet extended to the knowledge layer, and the gap is now the largest unspoken adoption blocker in the market. Dennis at Freshworks acknowledged the problem in unguarded terms: customers think their knowledge is current and accurate, and often it is not, and a customer success team must rearchitect and recurate the data before AI can produce any value at all. The acknowledgment is candid, and the gap exists across the service management vendor landscape, not only at Freshworks. Knowledge governance on most platforms today distinguishes between private and public and assigns role-based access; what it does not do is mark an article as a sensitive topic that AI may retrieve but only a human may respond to, or separate explanatory context from executable instruction within a single artifact. Knowledge governance designed for human readers, ported into an AI consumption layer without the architectural rework, will keep producing failures.

Let’s Connect

Have questions? That’s fantastic. Let’s connect and continue the conversation! Please reach out to me through social media or request a guidance session. Follow my blogs and research at Forrester.com.