How to Match the Right AI Brain to IT Finance

What you Need to Know Before Using AI in IT Finance 

“Curiosity killed the cat,” the saying goes.

In IT finance, curiosity is not the problem — unstructured curiosity is. Before you move from “I should probably look into AI” to applying it in your IT financial management, you need a clear view of what kinds of AI you’re dealing with and what they are (and are not) good at.

For IT finance, two families of models matter most right now:

• Large Language Models (LLMs)
• Large Quantitative Models (LQMs) 

In a recent video conversation with two fellow thought leaders, Nancy Braun, and Dr. Alexander Becker, we unpacked how these two types of models change the way we run IT finance — from reporting efficiency to forecasting accuracy and anomaly detection.

You can watch that discussion here: How to Match the Right AI Brain to IT Finance

This blog is the “orientation map” to that conversation. I’ll give you enough to get started on your AI journey — but I’ll leave some of the juicy details for the video and the whitepaper.

Two Different AI “Brains”: LLMs vs. LQMs

Think of LLMs and LQMs as two very different brains you can plug into your IT finance processes.

Large Language Models (LLMs) – the language brain 

LLMs are trained on tokens (words, code, and other textual patterns). They excel at:

• Understanding and generating natural language
• Summarizing and explaining complex material
• Creating narratives, emails, policies, and documentation

In finance, LLMs are increasingly used for report generation, narrative explanation, and self-service Q&A over financial data. But they have a big limitation: they are not calculators. Research shows that LLMs still struggle with precise numeric reasoning, especially over semi-structured financial data and complex math, and can produce plausible but wrong numbers.

Bottom line: LLMs are brilliant communicators, but you should not treat them as the system of record for your numbers.

Large Quantitative Models (LQMs) – the numbers brain

Large Quantitative Models are purpose-built AI systems optimized for numerical reasoning, forecasting, and complex data modelling. They’re trained on mathematical and financial datasets and are typically used for:

• Forecasting (e.g., IT spend, demand, capacity) 
• Scenario planning and optimization 
• Risk and sensitivity analysis 

Anomaly and Pattern Detection in Large Numeric Datasets 

You can think of LQMs as a far more sophisticated successor to traditional time-series forecasting and econometric models — but at a scale and complexity suited for today’s IT and cloud environments.

Until recently, most LQMs were custom-built inside banks, trading firms, and a handful of large enterprises. Now we’re starting to see commercial distribution of LQMs via cloud platforms, such as SandboxAQ’s quantitative models offered on Google Cloud, explicitly positioned as complementary to LLMs (Reuters)

For IT finance leaders it is important to know:  

• LLMs are increasingly easy to consume “off the shelf.” 
• LQMs are still more specialized, more data-hungry, and often require more investment, even as platforms begin to emerge. 

Where LLMs Make Immediate Sense in IT Finance 

You don’t need to rebuild your IT cost model to use LLMs. You can layer them on top of existing ITFM/TBM/FinOps tools and data to drive:

1. “Chat with your data”: self-service reporting & insight

LLMs can sit on top of your IT finance repository — allocations, service catalogs, GL, cloud cost data — and allow stakeholders to ask questions in natural language, such as: 

“Why did storage costs for our customer portal increase in Q3?” 

“Show me cloud spend per business unit vs. budget for the last 6 months.” 

“Which applications are driving the most unplanned infrastructure costs?” 

This does not replace your ITFM platform or data model — it exposes them in a more accessible way. 

Value: 

• Less time building one-off reports
• More time analyzing and acting
• Fewer “can you run this report for me?” tickets 

• Turning a complex variance analysis into an executive-ready summary 
• Tailoring explanations for different audiences (CFO, CIO, product owners) 
• Creating talking points for steering committees 
• Producing commentary for monthly IT cost reports or showback/chargeback decks 

• Drafting policies (e.g., cost reallocation rules, capitalization guidelines) 
• Generating and refining SQL or scripts for data extraction and transformation 
• Creating SOPs for month-end processes 
• Translating financial content across languages 

Used well, LLMs strip out mundane tasks, so your experts can focus on analysis, governance, and stakeholder engagement.

Where LQMs Shine: Forecasting, Quality and Early-warning

If LLMs let you talk to the numbers, LQMs reason with them. Here are three use cases: 

1. Modern forecasting and scenario planning 

LQMs are particularly powerful for predictive forecasting of IT spend across services, portfolios, and cloud resources. They can integrate:

• Historical IT costs (CapEx and OpEx) 
• Utilization and telemetry data (CPU, storage, network, licenses) 
• Business drivers (users, transactions, revenue, locations) 
• Market or external variables (FX, energy prices, vendor price lists) 

For IT finance teams, this means moving from simple “run-rate plus X%” approaches to dynamic forecasting that can: 

• Detect early trend breaks (e.g., new product launch, cloud migration pattern) 
• Simulate what-if scenarios: Examples: “What if we shift 30% of this application to a different cloud region?” 
• Support a “waterline” process — dynamically highlighting which services sit above or below agreed cost/performance thresholds and why. 

• Spotting misallocations or miscodings before they hit your reports 
• Detecting subtle anomalies that a rules-based system would miss 
• Validating data before it enters your ITFM or cost-allocation models 

Examples: Run vendor invoices, cloud bills, or asset feeds through an LQM-powered quality layer before loading them into the cost model — catching unexpected values, patterns, or missing fields. This not only reduces manual checking; it also improves trust in your IT cost data, which is the foundation for any meaningful AI initiative. 

Limitations you must respect 

Before you deploy AI against your IT budget, a quick reality check: 

LLMs – great with text, imperfect with numbers 

LLMs are strong at:

• Efficiency – drafting content, answering questions, exploring data 
• Translation – between languages, and between CFO-speak and CIO-speak 
• Interpretation – explaining complex material in human language 

However:

Be aware that there could be some hallucinations and mis-handle calculations and numeric context.

They must not be used as the source of record for financial statements, budgets, or allocations. 

They need guardrails for data privacy and regulatory compliance, especially with sensitive IT cost and vendor data. For more details take a look at CFA Institute Research and Policy Center 

Commercially, LLMs have a fast and easy onramp — embedded in many platforms and available via APIs — which is both their strength and their risk if governance is weak. 

LQMs – Powerful, But Not Plug-and-Play (yet) 

• Quality – anomaly detection and data validation 
• Speed – analyzing massive numeric datasets at scale 
• Prediction – forecasting and scenario analysis 

However:

They are not yet broadly available as off-the-shelf ITFM solutions; many are still custom-built or highly specialized.

They require strong data foundations (clean, well-modeled IT cost and usage data).

They need investment in skills (data science, MLOps, ITFM expertise) and clear model risk governance, especially in regulated industries.

The Real Opportunity: Combining LLMs + LQMs 

On their own, neither LLMs nor LQMs fully “solve” IT finance. Together, they start to look like an intelligent assistant that can both think in numbers and speak in business language. 

A few patterns we’re seeing:

LQM as the engine, LLM as the interface

• LQM forecasts IT spend, capacity, or risk. 
• LLM explains the results, tailors narratives, and proposes actions for different stakeholders. 

Bidirectional feedback

• The LLM helps you design scenarios for the LQM (“model these three cloud optimization strategies and compare the impact”). 
• The LQM returns scenario results; the LLM translates them into playbooks, recommendations, and board-ready materials. 

Bridge between IT and Finance

• LLMs can “speak both languages” — converting FinOps and TBM metrics into CFO-relevant views (and vice versa).
• LQMs then quantify the impact of decisions across IT services, business units, and financial periods. 

What you Should Know Before you Start 

If you are an IT leader or IT finance expert preparing to engage in AI, here are five pragmatic questions to answer upfront:

1. What business questions am I trying to answer? 
Example: “Where can I reduce run costs by 10% without harming key services?” 
Map those questions to LLM use cases (insight, explanation, productivity) and LQM use cases (forecasting, optimization, anomaly detection). 

2. Is my IT cost and usage data ready? 
Do you have a consistent ITFM/TBM model, service catalog, and cost allocation structure?  
Do you have cloud and infrastructure usage data with enough granularity for meaningful forecasting and anomaly detection? 

3. What guardrails do I need? 
Define what data LLMs can and cannot access (sensitive vendor contracts, personnel data, etc.). 
Establish human-in-the-loop checkpoints for anything affecting budgets, forecasts, or financial reporting.

Align with your organization’s model risk and compliance policies; regulators and professional bodies are already issuing guidance for AI in finance.See CFA Institute Research and Policy Center for more details.  

4. Which capabilities should I buy vs. build? 
Many ITFM/FinOps tools are embedding LLM capabilities for reporting, narrative generation, and guided analysis. 

LQMs may come via specialized vendors or cloud providers or be developed in-house for your specific IT cost drivers and business model (more details in our whitepaper).  

5. How do I organize the people side? 
Create a joint CIO–CFO–FinOps/ITFM working group.

Make sure you have the right mix of skills: finance, IT operations, architecture, data science, and governance.

Treat AI in IT finance as an ongoing operating capability — not a one-off project.

Where to Go From Here

This blog is intentionally a preview, not a playbook.

• The video touches on many of the above details such as the “waterline” process. 
• The whitepaper "AI for the IT Finance Practice" will unpack reference architectures, maturity stages, and a more detailed checklist for LLM and LQM adoption across ITFM, TBM, and FinOps practices, and how organizations like Siemens and others are approaching AI in IT finance. 

If you’re at the beginning of your AI journey in IT finance, your next step is not to “buy an AI,” but to decide which brain you need first, and where you can safely get early wins — while building the foundations for the combined LLM + LQM future.

And yes, stay curious — just make sure your curiosity is backed by clean data, clear use cases, and solid governance.