What is the strategy for running valuation (DCF/WACC, comparable or contingent) of an AI-led, Gen-AI led transformation to make it compelling for IT and business embracement. More I work with my CxO partners, industry leaders, start-ups and academia to propel adoption, I ponder !!

The 4 proponents of adoption

Pervasive curiosity across all industries, take for eg., tech services and operations, clinical trials, molecular research, curing nature’s imbalance “cross-pollination”, basics of concrete mix design optimization for carbon emission; leave apart intensive usage in space and military;

Ease of access and implementation of AI / model technology, which do not need specialized data science knowledge to deploy or train model – just APIs;

Availability of domain and technical talent to identify use cases and befitting solution;

Reducing labor cost by getting human like job done with acceptable precision, more bolstered by the now “Agentic AI” workflows and orchestration, taking decisions and moving alternate paths of end to end resolution of a task.

The hindrances of valuation and hence the ROI

What part of the valuation stay mysterious driving ROI determination fuzzy ? Yes certainly defining the “expense” (OPEX OR CAPEX)

Finding operating cash flow and cost impact year-on-year

Terminal value of growth or at maturity, on a multi-year north-star horizon, and then applying the “safest” company or industry DCF / WACC framework to find the economic value of such investment.

Compute cost – server, storage, back-up, for data pipeline, processing, business continuity etc – and it could be data centre driven CAPEX, or PaaS/SaaS/IaaS based Opex), or any combination of either

Energy spend – as part of your data centre infrastructure management strategy look at how much self generated energy CAPEX, or leveraging the grid as OPEX are you allocating or using;

Intangible legal regulatory cost of non-compliance due to imperfect use case, technology selection, testing process or at the least AI hallucination;

Business failure due to lack of right data strategy or again AI hallucination;

Liquidity risk of reduced debt / equity financing, due to achieving delayed or insufficient goals set

Competition risk of alternative MODELS, alternate business strategies an more

So what could be the line items of a potential earning statements look like where the CFO can probe, put alternate thoughts of optimization and find benchmarks ?

Where does that leave CIOs / CTOs / CFOs to accept as a strategy ?

Defining the industry benchmarks if it exists ?

Optimize strategy to search with LLMs leveraging precomputed indices, or resort to traditional search;

Use raw data science to optimize models;

Use smaller, fine-tuned models for queries, but it costs 50 % higher;

Implement techniques like model quantization and distillation to lower inference costs, accepting “risk” of accuracy;

Implement TLM, Trustworthy Language Model (TLM) — that flags incorrect LLM answers in real-time;

Drive a culture to accept results less robust or with less precision;

Run operations on intelligent programs during non-sensitive business time;

Use the investment for user facing processes – may be an agentic work flow, rather than build and train a foundation model.

It is just the beginnning

Execution and operational unpredictability are significant barriers to AI adoption for IT leaders but industry wants to “consume” AI / Gen-AI led capabilities for business users and IT operations ! Then what is the strategy of consumption to EXIT the death valley and generate economic value across the S-curve ? Stay tuned !!