As you may have noticed, things are rapidly changing in IT. The biggest change is the automation of tasks that used to be done manually. New frameworks such as DevSecOps and new Technologies such as AI are emerging, It probably is not discussed publicly that billions of Euros investments in legacy technologies/practices will vaporize in the next five to ten years. It is time to discuss the elephant in the room.

During research on IT Tools that are used to manage IT, I noticed that a number of frameworks co-exist that each have each introduced their own vocabulary and definitions. The same word in one framework has a different meaning in another framework. E.g. an “Asset” in ISO-27000 has a broader definition than an “Asset” in ITIL. Categorizing the tools by capability proved to be cumbersome, since the categories used are pluriform and overlapping.

Up to 2022, annually an overview was published which tools were used for which IT4IT Value Stream. If only one thing, that picture demonstrated the IT4IT landscape is becoming increasingly diverse and complex. The role that each tool plays in that landscape is changing, due to the ever changing way of working. The tools nowadays are linked to each other and workflows are chained, creating an end-to-end pipeline via which fine granular components/ configurations are developed. secured and operated. Tools that were used to manage fulfilment of human taks, are currently used rarely in the CI/CD pipeline.

ServiceNow holds a CMDB and an application/service portfolio, but often does not hold instances of fine granular components/configurations such microservices, containers and/or API’s that flow through the CI/CD pipeline. The content in the CMDB is up to date for things that have not changed in the last 24 hours. Ofcourse ServiceNow technically can instruct DevSecOps tools to test, integrate and deploy a configuration on the cloud, thus making ServiceNow the Orchestrator of Orchestrators. But I don’t know any actual implementation of this concept. ServiceNow also has an SCRUM/SAFe application onboard but it lacks the functionality and integrations that the competition is currently offering. Same applies to the Enterprise Architecture Capabilities of ServiceNow. DevSecOps tool vendors directly integrated their tools with each other and side-lined ServiceNow. ServiceNow’s licence policy/pricing is often less attractive than that of its competitors, albeit people often forget that (maintenance of the) integration of multiple point-solutions cost a lot as well. Anyway; when looking up an IT Landscape picture of DevSecOps tools and its integrations, it is noticeable the logo of ServiceNow is disappearing from recent DevSecOps Tooling landscape pictures. E.g. see this link, this link or this link. ServiceNow does not appear in Gartners Magic Quadrant for DevOps Platforms either.

BEST OF SUITE vs BEST OF BREED

Allowing DevOps teams to choose and integrate their own tool sets, has its advantages but also leads to a scattered tooling landscape, pluriform definitions, and lack of end-to-end visibility/control. Most companies also do not know what percentage of the DevOps team’s time is spend on maintenance of the IT tooling landscape (IT4IT costs are currently part of the overall IT costs). Those who ask around will find out that the true Cost of Ownership of (maintenance of the integrations of) the dozens of tools is significant. The relation between course granular IT solutions and fine granular IT components often is not well defined/maintained. Enterprise/Domain Architects, Release Train Engineers and DevOps Engineers, each appear to live in their own silo.

A number of large global enterprises hence decided to go into an opposite direction and use one Best or Suite Platform that loosely interfaces with multiple best of breed speciality tools. They have chosen to use the ServiceNow Platform to orchestrate all of its IT during its complete IT-Lifecycle. They accept ServiceNow is not the “best of breed” in many IT management areas, but overall probably still is the “best of suite”. They know that integration of dozens tools of dozens of vendors can provide the same (if not more) than what ServiceNow can provide, but (the maintenance of) such integration is neither simple or cheap. They use Application/Service Portfolio Management (APM) application to register the plans and ownership of course-granular IT-Solutions, use Strategic Portfolio Management to (SPM) to manage Demand, and Project Portfolio, use SAFe/AGILE to register/track the DevOps tasks, use IT Operations Management (ITOM) to discover/federate content for the CMDB, use IT Asset Management (ITAM) to keep track of what is where and used by whom, use Integrated Risk Management to manage Governance, Risk and Compliance, use SecOps to manage Security issues/breaches, and use IT Service Management to keep track of requests, questions, incidents, changes, and approvals.

The management of fine-granular digital components, and anything that needs to be managed/observed on real-time basis, is still done in technology-specific tools. ServiceNow needs to be integrated with those other IT4IT tools to make it work end-to-end. E.g. if a tool detects unavailability that cannot be automatically resolved, the tool may flag the outage to ServiceNow, where it may be determined which parties are/were impacted, and if action needs to be taken, by whom. Of course, for this, ideally the CMDB in ServiceNow is kept up to date on real-time basis, which is still rarely done.

Even though it is possible to cherry-pick certain ServiceNow applications, it is probably wise to first make a (broadly shared) decision to go for “best of breed” or for a “best of suite” strategy, and thereafter make a roadmap for (tools/practices in) the IT Domain, taking the market trends and emerging technologies into account as well. E.g. if one believes AI will kill SAAS within this decade, then one should probably not invest in frameworks that have an Return on Investment of more than 10 years.

Furthemore, before carving the strategy in stone, it is highly recommended to first investigate the prices of the different applications (and versions/capabilities thereof) of the ServiceNow platform. I.e. the IT modules ITSM, SPM, IRM, DevOps, SecOps, ITOM and ITAM each have their own licenses that must be paid per user/SKU/transaction. The total price per fulfiller per month, could easily be over hundred euros if a company procures Professional/Enterprise versions of all IT application licenses incl. AI support. The unit prices that apply for 36 months, typically are subject to a contractually committed unit volumes. If you are thinking to bring all data that you have into ServiceNow, you may first want to ask Sales the price per TB. That said, you can’t compare apples with pears. License prices are not the only/biggest portion of the Total Cost of Ownership. Having agility, velocity, consistency, and transparency within your Tooling landscape comes at a price.

Another “elephant in the room”: a decision for a best of suite approach, could easily lead to vendor-lock-in which may not be appreciated by one of the parties involved. On the other hand: with a best of suite approach, one can piggy-bag-ride on the product roadmap of the trusted suite-vendor. However, if the pig takes the wrong exit, or if the pig becomes greedy when you want to extend or expand, there is little you can do, unless you have mitigated those risks in the contract. E,g, if you foresee that automation will reduce the number of fulfillers that you will have/use in the future, you want want to agree in writing you can exchange fulfiller-user licenses for AI/SKU licenses, during the licensed period.

TRANSFORMATION TO THE FUTURE

Before this decade is over, AI Agents are expected to reason themselves, perform tasks autonomously, and adopt dynamically to what AI observes/experiences, on real-time basis. Rather than humans having to continuously train the Large Language Models, AI will train itself; it will become smarter, without requiring new technology. Maybe the AI-Agents will even teach each other new things, without any human intervention. The AI agents will directly interact with other AI Agents and with other IT Solutions of/for the Enterprise. The individual AI Agents will Create, Read, Update and Delete content in Databases.

When opening up the ServiceNow website on the home page it shows in big letters: “PUT AI AGENTS TO WORK FOR PEOPLE”. IMHO, this sentence summarizes the ServiceNow strategic direction. Vertical AI Agents will become the primary interface for people (being employees or customers of an Enterprise). Currently AI already can have a dialogue and create an intelligent response based on what it has been thought. Over time, the AI agents will become smarter as AI evolves from being a Large Language Model to genuine Artificial General Intelligence (AGI). The remaining dialogue for which the AI Agents can not yet generate a response themselve, will be summarized, put in a ticket and assigned to an appropriate specialized teams. The responses to customers regarding those tickets, most likely will be generated by AI.

It is not unlikely that multiple AI Agents will update information on real-time basis in a Configuration & Asset DataBase (CADB), where that data may be consumed by other AI Agents. That CADB probably will not be part of a SAAS solution; it will be a data lake that makes use of modern Databricks technology in which structured and unstructured data is made available to those that/who are authorised to that data. The content in CADB will be created and updated on real-time basis, by the tools that provision/commision IT artefacts and tools/clients that observe any changes in the IT infrastructure. It is not unlikely an AI-Agent will be developed that only has one purpose: Produce responses based on what is contained/changed/related in the CADB. The intelligence that can be obtained from the CADB will be used by any discipline that needs to know the status, dependencies, and whereabouts of any stack of Configurations or Assets that is relevant to the Enterprise (beit owned by the enterprise, or consumed -as a service- by the enterprise).

Some (including Microsoft’s CEO) believe, this decade we will see business logic that is statically programmed, will be replaced by by Business Logic that is dynamically generated and executed by AGI. By that time, AI will directly interact with data without making use of static application code. DevSecOps, SAAS, and low-code/no-code platforms “as we know it” will then disappear. I’m sure this will be the case for some (e.g SAAS) solutions within the next couple of years, but it will probably take longer time before all code will be replaced by AI. E.g. a lot of institutions, today are still using cobol code from the eighties. When AI can interpret and convert legacy code into Intelligence itself, things could transform really fast.

Zooming in on the Enterprise Support Domain:

It is blatantly clear, AI will gradually take over a significant portion of work that is currently being done by helpdesk staff/fulfillers and static workflow tools. The dialogue between AI and people will remain, and IT-work and other Enterprise support-work still needs to be done, albeit (by the end of this decade) probably no longer will be done by humans, but by AI Agents and AI-Robots.

Zooming in further on ServiceNow’s role in the Enterprise Support Domain and in the Customer Support Domain: ServiceNow currently is a large dominant player in a niche ITSM/CSM SAAS-market, which SAAS market will likely largely disappear over time. There are three scenarios which will allow ServiceNow to remain successful as a company in the upcoming decades:

  • ServiceNow can try to create a new “Enterprise AI Support” niche market and become the dominant player in that space, by allowing its customers/partners to create specialised vertical AI-agents themselves that directly engage with Employees and Customers and that interface with traditional Enterprise-IT (SAAS) solutions via microservices/API’s using the technologies of the big AI providers.
  • ServiceNow can (and probably will) also dive into the niche of (external) Customer Service Management and Field Management for large Enterprises, where ServiceNow currently has little competition and future use of AI Agents is a no-brainer.

If ServiceNow wants to survive in world that rapidly gets automated, it probably needs to go “all-in” to AI, and do it now, before newcomers start taking over ServiceNow’s clients.

It looks like that is exactly what ServiceNow is doing. ServiceNow is being integrated with multiple AI vendors and wil support a “Bring Your Own Model”, rather than becoming an AI vendor themselves. By 2027 (2-3 years from now), ServiceNow predicts the AI Agents on the platform to be fully autonomous. It’s a bold transformation strategy, but most likely the only sustainable strategy on the long run, when the current traditional SAAS revenue streams will have dried up.

What about the People and the Processes?

Although not many companies have embraced it yet, the technology is already there that can engage with people and that can automatically generate intelligent responses without any human intervention. The responses currently are limited to what the model has been thought. That said, many IT service desks that are contacted via mail, portal or phone, often are “non-skilled” and provide basic IT support in a limited number of languages, before summarizing the dialogue (in english) and assigning a ticket to a specialist group.

In 2025, if not done already, companies will train their LLM’s and start offering AI as a first point for contact for their employees. The engagement with the employee (in one of the many supported languages) is captured, summarized and stored. If still needed, a ticket can be assigned to a helpdesk or to a specialist resolver group. Companies that have their application/service portfolio incl. commitments and entitlements in order, can feed that info into the language model, so that tickets can automatically be assigned to the (then) correct specialist team. The more intelligent AI becomes, the less value a Help Desk can add, especially if AI can also talk/chat to the employee in their own language. Bi-directional translation of the dialogue is done on real-time basis; knowledge/experience kan be entered in any language and is presented to people in their own language.

Any call that does not have to go to an Enterprise Service desk, saves a company 10-40 Euro per ticket. On average, each employee raises 20 support tickets per year. Companies with 50K employees thus spend 10-40 million euro per annum on their Enterprise Service Desk(s). A 25% reduction in the number of tickets would already be a saving for them of 2.5 to 10 million euro per enterprise per annum. Companies such as wal-mart have 2+ million employees and can save 100-400 million dollar per annum if they implement AI in 2025. That said, the implementation and use of AI definitely is not free of charge. The license cost of a Standard Platform (that does not have support for AI) is 50% cheaper than a Pro Platform without AI. PRO platforms that also offer AI capabilities are approx 30% higher than that of a PRO platform without AI. A Transition from Standard license w/o AI to a PRO platform that includes AI, could easily double the license fees. Note: actual dollar figures may differ based on contracted volumes, and capabilities.

The same AI technology can also be used to assist service desk agents and specialist fulfillers to resolve tickets that have been assigned to them. AI can inform the employee of progress and resolution in his own language whilst the assignee could be speaking chinese only. If the LLM’s gets fed the information in resolved incidents, AI’s knowledge will automatically increase and less incidents still have to be assigned to teams. This will improve/extend the incident process (which is useful for as long as the manual process is still needed) and will replace the knowledge management process that many companies use today.

As a logical next step, AI agents will do/trigger things themselves. The most obvious/simple use case for this, is an AI agent that automatically resets a(ny) password for a user if so requested by the user. A slightly more sophisticated AI agent can timely order a pallet of new PC’s if it detects spare-stock is almost running out. An email address and a user- identity can automatically be created by the AI Agent, if the AI agents detects (or is informed) a new employee is joining the company.

There are hundreds of use-cases that can be fulfilled by creating AI Agents for it. This will probably keep the current IT community quite busy until the next generation of AI capabilities will emerge (in 2026?). All of the above work, that used to be done by people, and that will be automated by AI, will disappear including the processes that were used by the people. That said, AI Agents (for now) still need to be created and maintained by people. However, don’t get your hopes up; in due time there will be AI Agents that will generate (updates of) AI Agents on near-real-time basis. By that time, all people will play golf daily, on golf courses that are continuously maintained by AI robots :-)