ITIL 5, AI, and the return of grown-up service management

Barry Corless considers why accountability, governance, and judgement matter more than ever in an AI-driven world

This piece isn’t an argument for ITIL 5, AI, or any particular operating model. It’s observational and drawn from conversations I’ve had in conference halls, in steering groups, and even late-night incident calls. The characters are fictional, but the viewpoints are not.

If you’ve spent any time in IT, you’ve heard all of these voices before. Often in the same room; sometimes even in your own head! What follows isn’t about taking sides, but it’s about recognising the tensions that keep resurfacing as automation and AI creep closer to the heart of evolving service management. Read it not as advocacy, but as a mirror.

If it feels uncomfortably familiar, then I’ve achieved what I set out to do.

The cast:

• The Pragmatist
  Was taught ITIL v2, then built their career on ITIL v3 and believes in control, documentation, and stuff that works at 3am. Had smart appliances networked at home before you could say “Alexa”.
• The Agilist
  Survived ITSM and still bears the scars! Thinks frameworks slow delivery and confuse accountability and last attended a CAB in 2011. Has a vast collection of laptop stickers and socks from DevOps conferences.
• The AI Vendor
  The one who smiles a lot. In any conversation, they mention Copilot and ChatGPT within 30 seconds. Believes in their product and will happily sell their granny if she’s got a compatible API!
• The Wise One
  Has seen enough transformations to know that extremes always fail. Brings balance to the force. Occasionally sighs and issues reminders of winning the Paul Rappaport Award for outstanding contribution to ITSM many, many years ago.

The four are former colleagues who meet up around the dinner table at a random conference. Inevitably talk turns to ITSM and the release of ITIL 5.

SCENE 1: RE-OPENING OLD WOUNDS

Pragmatist:
Before anyone starts, let me be clear that ITIL v3 worked. It wasn’t pretty, but that service lifecycle worked. Since then, all I’ve seen is abstract diagrams, and people talking about “value” without fixing a single incident faster.

Agilist:
Ah! That’s because ITSM was never the answer, was it? Heavy processes, handoffs everywhere, change boards killing flow. We escaped all that for a reason. At least I did…

AI Vendor:
Interesting points you both make. Now, imagine all of this solved with AI! Autonomous incident resolution, predictive change, self-healing infrastructure. We can deploy in six weeks and… Sorry, I’m slipping into work mode there.

Pragmatist:
Six weeks to do what, exactly?

AI Vendor:
Transform service management.

As they exchange telling “we’ve been here before” glances around the table, the Wise One interjects…

Wise One:
Right, let’s slow this down before anyone transforms anything.

SCENE 2: CUTTING TO THE CHASE

Wise One:
You’re all arguing past each other because this isn’t about frameworks or agility or AI. It’s about who is in control when things go wrong or need changing.

Agilist:
Easy, teams should be in control.

Pragmatist:
But someone – an individual – must be accountable, surely?

AI Vendor:
Of course, the AI system learns that…over time. (pauses) And don’t call me Shirley.

After a 2-minute interlude talking about scenes from Airplane…

Wise One:
And there it is! Three answers with one common problem.

SCENE 3: THE PRAGMATIST SPLITS UP WITH ITIL 4

Pragmatist:
I’ll tell you why people like me didn’t trust the last evolution of ITIL. The ITIL that must not be named. It stopped talking about control and started talking about feelings. Co-creation, journeys, value streams and meanwhile outages still needed fixing at 3am. Co-creation never bounced a server when your production systems were lying flat on their back.

Wise One:
That’s a fair cop. I guess it felt like the discipline evaporated?

Agilist:
Ah, but… from my side, it still felt like bureaucracy with nicer words. The CAB disappeared from the renamed Change Evaluation practice, but nobody took any notice. They still ran CABs. I just never went!  

Wise One:
Exactly, one of you felt control was diluted. The other felt nothing really changed. Neither of you really engaged with ITIL 4. But, you were both still using ITSM practices in one form or another.

SCENE 4: AI. THE TROJAN HORSE?

AI Vendor:
Which is why AI is different! It removes flawed human error and emotion from the loop.

Pragmatist:
You mean it removes human responsibility and accountability. Before we all get gooey and misty-eyed about autonomous AI, do yourselves a favour and study what Hollywood has taught us. Computers are the villains. The WHOPPER inWar Games; Skynet in The Terminator; and VIKI in I, Robot. Every single one of them started with the same promise: “remove human error, optimise decisions, act faster than people ever could.” Every single one of them went wrong because humans stepped back and assumed the machine “knew best”.

Wise One:
I get it. Now, I hope you’re not saying Agentic AI will start a nuclear war? I do hope you are saying that if you design systems without clear human accountability, you eventually lose the ability to stop them when they’re ‘confidently wrong’.

Silence.

Wise One:
AI doesn’t fix bad data, unclear ownership, or inconsistent processes. It scales them… quickly, confidently, and repeatedly with the world cheering on.

SCENE 5: THE RETURN OF THE HUMAN… WITH CONTROLS

Agilist:
So what, we centralise everything again?

Wise One:
No. We stop pretending autonomy means everyone invents their own definitions, rules and practices.

Pragmatist:
Now you’re talking.

Wise One:
Federated governance is simple when you strip the ideology away: one set of definitions; one set of data standards; one set of non-negotiable policies. Then everything else is local judgement within guardrails.

Agilist:
So teams still decide how they work?

Wise One:
Yes, but they play by a few rules. They don’t get to redefine what an incident is, and they might need to adhere to common formats for universally accessed data or have pre-defined boundaries for testing changes based on organisational need.

SCENE 6: THE CASE FOR THE LOWER MATURITY ORGANISATION

AI Vendor:
But you’re limiting innovation. I spend so much time working with tens of lower maturity organisations who just can’t wait to get their hands on AI.

Wise One:
Listen, I’m not limiting innovation; I’m limiting the blast radius from ‘Hollywood AI Thinking’. Poorly thought-out AI isn’t an answer.

Pragmatist:
I’ve always liked you, Oh Wise One. I can see why you won that award now!

Wise One:
Did I mention I’d won…? Seriously though, in organisations barely ready for automation, AI should:

• Recommend, not decide
• Assist, not replace
• Explain, not obscure

If you don’t understand a process enough to automate, you have no business letting AI improvise it.

Agilist:
So AI proves itself as a junior consultant before becoming a principal and a VP?

Wise One:
Exactly.

SCENE 7: ITIL 5. THE FRAMEWORK FOR PEOPLE WHO HATE FRAMEWORKS?

Pragmatist:
So, where does ITIL 5 fit into all this?

Wise One:
Nowhere… if you present it as a revolution.

Agilist:
Go on.

Wise One:
ITIL 5 will work if it’s framed as:

• Keeping what works – it’s a guiding principle in ITIL 4 and ITIL 5
• Dropping what doesn’t – particularly if it doesn’t add value. That’s another guiding principle retained between ITIL 4 and 5
• Tightening accountability and ethics in an automated and AI-driven world – that’s a whole new section of the guidance

Most people have adopted an ITSM framework to an extent, now we must stop re-arguing the basics and run the place properly.

Pragmatist:
So it’s less ideology, more the maintenance manual.

Wise One:
Precisely.

SCENE 8: THE ENDING NOBODY EXPECTS…

AI Vendor:
So… where do we deploy AI first?

Wise One:
Hold your horses! You can use it in a limited sense to help create and curate data but please no heavy lifting with it until it learns!

AI Vendor:
But, but, but…

Wise One:
You fix your data and governance. What decisions do we need to make? And why? I’d prefer to make decisions evidence-based, too. So, standardise your records; agree your decision rights; automate the boring stuff. Then, and only then, you invite AI into the room to orchestrate and decide. It’s learnt what’s right and wrong with data you trust. Ergo you can trust its decisions more confidently.

Agilist:
That’s… err… annoyingly sensible.

Pragmatist:
It feels like ITIL 3, but without the dead weight.

Wise One: (staring wistfully into an almost empty wine glass)
I used to think frameworks were the only answer. That if organisations failed, it was because they didn’t believe ‘hard enough’, or they picked the wrong horse and rode it badly. My experience cured me of that. You know that frameworks don’t fail because they’re wrong. They fail because people expect them to think on their behalf.

And AI? AI won’t replace judgement. It will expose where judgement never existed in the first place. Giving AI decision-making authority without validated data-based judgement is like building another Death Star because the rebels can’t possibly find a weakness for a third time!

The real story is that ITIL 5 won’t matter if it becomes just another belief system. It will matter if it helps organisations remember four simple truths:

• Humans still own decisions.
• Intelligent Automation comes before Artificial Intelligence.
• Governance exists to protect good judgement, not suffocate it.
• ITIL 5 is a reference model, not a replacement for thinking.

Get that right, and nobody will argue about frameworks anymore. They’ll just notice that things work.

Dessert arrives and the conversation turns to their old CIO who recently retired…

A FINAL THOUGHT

If ITIL 5 and AI are to succeed together, it won’t be through blind obedience.

They will succeed by being treated as what they are: a maintenance manual and power tools, used by professionals to:

• Keep the lights on
• Keep humans accountable, and
• Prevent the next ‘shiny thing’ from breaking production.