IT forecasting is familiar territory for everyone in IT finance: Budgets are created once (twice) a year. Variances are reconciled after the fact. Spreadsheets and “tribal knowledge” fill the gaps. That process has served us for years. But it is also backward-looking, labor-intensive, and increasingly misaligned with how digital, cloud, and AI services behave.
In this fourth blog of our AI-Powered IT Finance series—based on the whitepaper “AI for the IT Finance Practice” and our final video—I zoom into a concrete use case:
How can AI transform IT forecasting from rear-view reporting into an early warning system?
To deep dive into the use case, let us look at the current problems related to IT forecasting:
You do not jump from Excel-based forecasting to AI-powered IT forecasting overnight. In practice, most organizations move through three overlapping phases.
Step 1 – Use LLMs to relieve the staff squeeze and “shift left” into analysis. You may still forecast using traditional methods, but you can dramatically change how you get there. Large Language Models (LLMs)—already embedded in many FP&A and ITFM tools—are excellent at turning messy inputs into cleaner drafts (reports, commentaries, decks). These tools can answer ad hoc questions in natural language (“Why did storage spend spike last month?”). Additionally, they help create narratives for different stakeholders (CFO vs CIO vs product owner).
What this means: You still own the numbers, but AI accelerates how you prepare, explore, and explain them. That is your first “shift left”: less time building reports, more time questioning them.
Step 2 – Bring in LQMs for smarter, driver-based forecasting. Once you have a solid ITFM/TBM model and at least 2–3 years of consistent historical data, you can go further. Now, Large Quantitative Models (LQMs) and other machine learning models become your “numbers brain”. They can combine: historical IT costs (Capex & OpEx), utilization & telemetry (CPU, storage, licenses, tickets), business drivers (users, transactions, locations, revenue) and external factors (FX rates, energy prices, vendor price lists) to deliver predictive, driver-based forecasts that go beyond simple run-rate trending.
What this means: Forecasting becomes faster, richer, and more objective, grounded in the actual behavior of your services and demand, not just historical averages.
Step 3 – Leverage AI agents as financial co-advisors. The next step (and where many vendors, including Serviceware, are heading) is agentic AI: AI agents that orchestrate LLMs, LQMs and your ITFM data in a continuous loop. Imagine an AI financial advisor for IT that:
Continuously monitors for example your ITFM data, cloud spend, and business drivers. Uses LQMs to update short- and medium-term forecasts automatically. Uses an LLM to interpret the results and flag what matters, in plain language. Or proactively alerts you to specific trends e.g. “If this trend continues, Service X will exceed its annual budget by 18% in Q3.”
What this means: Forecasting becomes more like a living conversation between IT, Finance, and AI—where anomalies and risks surface earlier, and your people focus on judgment and negotiation, not detection.
People’s skills and mindset must be upskilled. You need more than a generic “What is AI?” training. Your teams need hands-on, role-specific enablement around AI-specific skills: How to use AI-enabled features in your ITFM/FinOps tools. Understanding model outputs: forecasts, confidence intervals, anomalies. Basic “prompting” skills: asking the right questions of LLMs. But also, around soft skills. You need to address the change in mindset from “AI as threat” to “AI as copilot,” comfort with experimentation and iterative pilots, cross-functional collaboration across IT, Finance, and data teams.
Establish a process and data to give AI something trustworthy to work with. AI does not fix weak data or missing structures. It amplifies whatever you give it. For forecasting, you’ll need: 2–3 years of consistent IT cost data mapped to services, towers, and consumers, operational and business drivers: tickets, users, transactions, locations, project portfolios, revenue proxies, external inputs: vendor price lists, FX rates, IT service cost benchmarks from Serviceware, industry benchmarks from sources like IDC, Gartner, and others.
You also need clear processes for: When and how forecasts are refreshed (monthly, rolling, scenario based), how anomalies are reviewed and escalated, how AI-generated insights are documented and approved within your financial governance.
Automate it with an AI-powered ITFM solution. Finally, you need a platform that brings data, models, and collaboration together. An AI-ready ITFM platform should provide a single, structured model for IT costs, services, and consumers leveraging DVM, TBM, FinOps frameworks, support planning & forecasting at different levels (service, tower, BU), integrate with cloud cost data, operational systems, and business metrics, and offer hooks for LLM-based assistants and LQM/predictive models. Start the automation journey with an AI-powered ITFM solution.
Serviceware COO Alexander Becker provides a sneak peek at what Serviceware’s ITFM can offer (watch more in the video):
🔮 Predictive forecasting with human-assisted AI
🚨 AI-supported anomaly detection
Finally, do not forget governance as AI in finance must be controlled.
Regulators and standard setters are clear: AI in finance brings model risk, data quality challenges, and systemic risk if poorly governed.
For IT finance, that means:
End results to look for (aka: knowing when you have reached your goal). The shift in IT forecasting will be from “What happened?” to “What’s likely to happen—and what should be done about it?”
👉 Watch the video on AI-Powered IT Finance for the full discussion and live examples.
📄 Download the whitepaper “AI for the IT Finance Practice” to explore reference architectures, maturity stages, and case studies.