Meet VERA, My AI-based value-investing engine.

Fundamentals 35. This article explains the concept of VERA, short for Value-investment Equity Research Agent, my AI-based value-investing engine.  

Over the past few months, I have been developing an AI-based fundamental analysis tool called VERA, short for Value-investment Equity Research Agent.

The goal was to have an AI assistant that could help me carry out a fundamental analysis of a company. I am still fine-tuning VERA. But it has reached a stage where it can produce a first draft of a company’s fundamental analysis.

Of course, I still have to edit it so that it reads better. But the gist of the analysis is there. Rather than wait for perfection, I decided to start publishing the various edited versions on my blog so that I can get the appropriate feedback.

Each article aims to answer two questions: Is this a good business? And is it worth buying at today’s price?

If you have ever read an equity research report and wondered how an analyst moves from financial data to a BUY, HOLD, or SELL conclusion, these articles offer a look at that process. 

VERA follows a disciplined analytical framework, applying it consistently and making the reasoning visible for readers to examine and challenge.

Contents

• What VERA actually is

• From a ticker to an article: the journey

• Why you can trust the numbers

• What VERA is not

• How I will use VERA on this blog

This blog is reader-supported. When you buy through links in the post, the blog will earn a small commission. The payment comes from the retailer and not from you. Learn more.

What VERA actually is

VERA is a software analyst. You give it a stock ticker - say, a technology company or an industrial giant - and it returns a long-form investment article: several thousand words of narrative that reads like something you would find on SeekingAlpha or ShareInvestor, complete with a reasoned conclusion. 

It is not a screener that spits out a score, and it is not a chatbot improvising an opinion. It is a fixed, defined analytical process that has been encoded step by step, so that the same company examined on two different days is examined the same way.

That last point matters more than it sounds. A great deal of what passes for AI investment commentary is inconsistent: the analyst emphasises growth for one company and the balance sheet for another, depending on what tells the more flattering story. 

VERA does not get to choose. It runs every company through the same sequence of questions, applies the same tests, and lets the conclusion fall out of the evidence rather than deciding the conclusion first and hunting for support.

The philosophy: good businesses at fair prices

VERA is built on the tradition of value investing - the school of thought associated with Benjamin Graham, Warren Buffett, and the valuation work of Professor Aswath Damodaran. 

The core belief is straightforward. A share is not a lottery ticket; it is a fractional ownership of a real business. Over the long run, the price of that share should track the value the business actually creates. 

So, the job of analysis is to estimate what a business is genuinely worth - its intrinsic value - and then compare that to the price the market is asking today. Two ideas follow from this. 

• First, quality matters: a wonderful business bought at a fair price will usually beat a mediocre business bought cheap. 

• Second, price matters: even a wonderful business can be a poor investment if you overpay. 

VERA is designed to weigh both - to ask ‘is this a good business?’ and ‘is this a fair price?’ - and to insist on a margin of safety, a discount to intrinsic value that protects you when, inevitably, some of the assumptions turn out to be wrong.

This philosophy is organised into four pillars, which are the internal scaffolding of every analysis VERA produces: 

• Absolute analysis - is the business good on its own terms?

• Relative analysis - how does it compare with its rivals?

• A financial overlay - is the balance sheet sound?

• Valuation - is the price fair, and with what margin of safety? 

The rest of this article walks through how VERA moves from a bare ticker to a finished verdict across those four pillars.

👉 To know a bit more about the architecture and design approach, refer to the Appendix - subscribers can access the complete password-protected Appendix.

👉 Subscribers only. Click here and enter your access password. New here? Sign up to receive your free access password.

From a ticker to an article: the journey

Step 1 — Start with real numbers

Everything begins with data, and VERA is strict about where that data comes from. It pulls audited financial statements - income statements, balance sheets, and cash-flow statements going back roughly a decade from established financial-data providers and directly from companies’ regulatory filings. 

Crucially, it never invents a number. If a figure cannot be sourced, VERA does without it rather than filling the gap with a plausible-looking guess. Every data point it uses is tagged with its origin the moment it is collected, so any claim in the final article can be traced back to a real statement.

This sounds obvious, but it is where a lot of automated analysis quietly goes wrong. Models that are allowed to ‘estimate’ missing inputs can drift a long way from reality without anyone noticing. VERA’s rule is the opposite: real data or nothing.

Step 2 — See the true economics, not just the accounting

Reported accounts and economic reality are not always the same thing, and this is one of the areas where VERA earns its keep. 

Accounting rules sometimes flatter a business and sometimes punish it, and a naive reader of the headline numbers can be badly misled.

Take a fast-growing software company that spends heavily on research and development. Accounting rules force it to treat that R&D as an expense the moment it is spent, which can make a genuinely profitable business look like it is losing money year after year. 

But that R&D is really an investment that pays off over several years - much like building a factory. VERA recognises this: it capitalises the R&D, spreads the cost over its useful life, and recalculates the company’s profits and returns on that economic basis. The result can flip a business from an apparent loss-maker into what it truly is - a profitable enterprise investing aggressively in its own future.

VERA also watches for what it calls structural breaks - moments when a company fundamentally changed shape. When a firm makes a transformational acquisition or sells off a major division, the years before and after are effectively two different businesses. 

Comparing them as if they were one produces nonsense: a growth rate that is really just the accounting artefact of a disposal, for instance. VERA detects these breaks and scopes its analysis to the business as it exists today, so it is measuring the company you could actually buy, not a statistical blend of the company that used to be.

Step 3 — Is this a good business?

With the economics straightened out, VERA asks the central question of the first pillar: is this a good business in absolute terms? It assesses this across four dimensions drawn from the Damodaran valuation framework - growth, profitability, reinvestment, and risk.

Growth: Is revenue expanding, and is that expansion coming from the business winning more customers (organic growth) or simply from buying other companies (acquisitions)? The distinction matters enormously - organic growth is usually far more valuable and durable than growth bought with shareholders’ money.

Profitability: This is the heart of it. The key test VERA applies is whether the company’s return on invested capital - the profit it earns on every dollar tied up in the business - exceeds its cost of capital, the return investors require for taking the risk. 

A business that earns more than its cost of capital is creating value with every dollar it deploys. A business that earns less is quietly destroying it, no matter how fast it is growing. This single comparison - returns versus the cost of capital - is arguably the most important line in any VERA analysis.

Reinvestment: When the company ploughs money back into itself, does that spending earn good returns, or is it pouring capital into projects and acquisitions that do not pay off? Great businesses are great precisely because they can reinvest large sums at high rates of return for a long time.

Risk: How financially fragile is the company, and how predictable are its results? Risk carries disproportionate weight in valuation, because uncertainty about the future is exactly what discounts a business’s worth.

Alongside these, VERA looks for a moat - a durable competitive advantage such as high switching costs, a powerful brand, network effects, or a cost advantage that rivals cannot easily copy. 

A moat is what allows a good business to stay good; without one, high returns tend to attract competition and fade. VERA does not just assert a moat exists - it checks whether the financial evidence (sustained high returns, stable margins) actually supports the claim, and it says so plainly when the evidence is thin.

Step 4 — How does it stack up against rivals?

A business can look impressive in isolation and still be the weakest player in its industry. 

So the second pillar is relative analysis: VERA lines the company up against a handful of genuine peers and compares them on the same footing — revenue growth, returns on capital, profit margins, cash generation, leverage, and earnings growth.

The point of this exercise is not to crown a winner but to understand context. A company might earn a respectable return on capital yet still trail every one of its competitors, which tells you something its own numbers cannot. 

VERA is careful here to compare like with like, and - importantly - it does not pretend to know why a rival performs as it does when it has only analysed the subject company. It reports where the company stands and what that standing implies, and stops there.

Step 5 — Is the balance sheet strong enough?

The third pillar is a financial overlay: a hard look at the company’s financial health. 

• Can it comfortably service its debt? 

• Does it generate real cash, or only accounting profits? 

• Would it survive a bad year? 

This layer acts as a modulator on everything else. Strong finances amplify a good business’s prospects; weak finances can turn a promising story into a dangerous one, because a company that cannot fund itself through a downturn may never get to realise its potential. 

VERA grades this soundness on a consistent scale rather than by gut feel, and - as I have learned while testing it - that grade is deliberately conservative about leverage, because debt is the single factor most likely to turn a disappointment into a disaster.

Step 6 — What is the business actually worth?

Only now, having decided whether this is a good business, does VERA turn to price. 

The fourth pillar is valuation, and VERA’s primary tool is a discounted cash-flow model - the standard method for estimating intrinsic value. In plain terms, it projects the cash the business is likely to generate over the coming years, then discounts those future cash flows back to today’s money using the company’s cost of capital. 

Cash arriving ten years from now is worth less than cash in hand, and the discount rate captures exactly that. Add up all those present-day values, and you have an estimate of what the whole business - and therefore each share - is intrinsically worth.

VERA states its assumptions openly: the growth rate it expects, the profit margins, the reinvestment, the discount rate. 

This is deliberate. A valuation is only as good as its inputs, and by laying them out, VERA lets you see exactly what it assumed and disagree if you wish. There is no black box producing a single magic number; there is a chain of reasoning you can follow and challenge.

The intrinsic value is then set against the market price. If the shares trade well below intrinsic value, there is a margin of safety - a cushion that protects you if the future proves less rosy than the model assumes. 

VERA does not use a one-size-fits-all cushion: it demands a bigger discount for shakier businesses and accepts a smaller one for proven compounders. A company can be excellent and still not a buy today simply because the price already reflects all the good news.

Some businesses cannot be sensibly projected - a young company that has only recently turned the corner, for example. 

For these, VERA flips the question around. Instead of asking ‘what is this worth?’, it asks ‘what would have to be true to justify today’s price?’ - working out the growth and profitability the market is implicitly betting on, and then judging whether those expectations look reasonable against the company’s own history. 

This reverse-engineering is often more honest than forcing a precise valuation onto a business that does not yet support one.

Step 7 — Turning analysis into a story

Finally, VERA writes. But it does not write a scorecard or a bullet-pointed checklist. 

It produces a flowing article that opens with a genuine tension or paradox - the thing that makes the company interesting or difficult - and then walks through the evidence so that the conclusion emerges naturally at the end, the way a good essay builds to its point. 

The BUY, HOLD or SELL is never announced in the first line and then defended; it is earned across the argument.

The article moves through the company’s background, its operating performance, the economics of each sale, its financial position, how it compares with peers, its prospects, what the market is currently pricing in, the intrinsic-value calculation, the main risks, and a conclusion. 

Every quantitative claim in that narrative is one that can be traced back to the underlying data. The aim is a piece a thoughtful investor could read, understand, and act on - without needing a finance degree to follow it.

 

Why you can trust the numbers

There is a well-known weakness in using AI for anything numerical: language models are fluent, but they are not calculators, and they can state a wrong figure with complete confidence. 

VERA is built specifically to neutralise this. All of the arithmetic - the growth rates, the returns on capital, the discounted cash flows, the margins - is computed in ordinary, auditable program code, not by the AI. 

Those verified figures are then handed to the writing stage as fixed facts that must be used exactly as given. The AI’s job is to explain and narrate, not to do the sums. If it ever tries to slip in a number of its own, automated checks strip it out.

On top of that, the whole system is wrapped in an unusually thorough testing regime. Over many months I have run VERA against a set of deliberately awkward reference companies - a cash-rich software firm, a transformed industrial conglomerate, a business carrying heavy goodwill, a loss-maker in turnaround - precisely because they break naive analysis. 

Each of those cases has surfaced subtle errors that a clean-looking test suite missed, and every one has been fixed and locked down with a new test so it cannot return. The discipline I hold the tool to is simple: passing tests is not proof; the proof is running the full analysis end to end and checking that the output makes sense.

When VERA is uncertain, or when the data is anomalous, it says so rather than papering over it. A strange tax rate, a data point that looks unreliable, a company whose reported capital base is distorted - these are flagged in the analysis, not silently smoothed away. Honesty about what is not known is part of the design.

What VERA is not

It is worth being equally clear about the limits. VERA is not financial advice, and nothing it produces - or that I republish here - should be taken as a recommendation tailored to your circumstances. It is one disciplined opinion, generated by software, offered as a starting point for your own thinking.

VERA depends on the data available to it, and that data has boundaries. Coverage is strongest for larger, widely-followed companies; smaller or more obscure names may be out of reach. 

Figures are only as fresh as the last time the underlying sources were updated, so a fast-moving situation can outrun the analysis. 

And a valuation, however carefully built, rests on assumptions about an unknowable future - VERA states those assumptions precisely so you can judge them, but no model can promise they are correct. 

Finally, every article is transparently attributed as AI-generated. VERA does the heavy, consistent, unglamorous work of building the case; the judgement about whether to act on it remains yours, and mine.

How I will use VERA on this blog

I plan to first use VERA to run a company analysis. This is then reviewed by me. 

The goal is not to tell you what to buy. It is to model a way of thinking about businesses -patiently, quantitatively, and with a healthy scepticism about price - that I have found valuable in my own investing. 

Some weeks the verdict will be enthusiastic; more often, I suspect, it will be a reminder that a fine company is not the same thing as a fine investment when the price has run ahead of the value.

Think of each write-up as a worked example: here is a real business, here is the evidence for and against it, here is roughly what it seems to be worth, and here is why the conclusion lands where it does. 

If, over time, these pieces sharpen how you read a company’s numbers and how you think about paying a fair price for quality, then VERA - and this blog - will have done their job.

One last thought on why I built VERA this way rather than simply asking a chatbot for its view on a stock. 

An off-the-cuff answer is only as good as the day it was given and the mood of the model that gave it; ask twice, and you may get two different stories. 

What I wanted instead was a method - something repeatable, inspectable, and stubborn about its own rules. 

VERA will not flatter a company because its story is exciting, and it will not condemn one because its recent numbers look ugly, if the underlying economics say otherwise. 

It shows its working at every step, states the assumptions it cannot avoid making, and hands the final judgement to a human who can weigh things a model cannot: management character, industry shifts, and the simple question of whether a thesis still makes sense in the real world. 

That division of labour - the machine for the discipline and the arithmetic, the human for the judgement - is, to me, the honest way to use these tools. I hope the weekly pieces show it in action.

END

Disclaimer & Disclosure

I am not an investment adviser, security analyst, or stockbroker.  The contents are meant for educational purposes and should not be taken as any recommendation to purchase or dispose of shares in the featured companies.   Investments or strategies mentioned on this website may not be suitable for you and you should have your own independent decision regarding them. 

The opinions expressed here are based on information I consider reliable but I do not warrant its completeness or accuracy and should not be relied on as such. 

I may have equity interests in some of the companies featured.

This blog is reader-supported. When you buy through links in the post, the blog will earn a small commission. The payment comes from the retailer and not from you.

Your link text