Seeking Input on an Investment Thesis Based on “AI Readiness”
After examining a narrow but illustrative segment of the U.S. economy, it appears that many large corporations are far less prepared for AI adoption than current market narratives suggest. I recently outlined a concrete example in the following thread:
https://www.wallstreetoasis.com/forum/off-topic/what-ai-revolution
Synopsis:
Most major fast-food brands have been failing online-order customers for nearly a decade by not implementing relatively simple AI-driven forecasting models—models that have existed for years and can run on modest hardware. Specifically, order prep-time estimates often rely on a narrow slice of available data (only past busyness averages) and fail to incorporate real-time factors such as current order queues or staffing levels.
This failure could stem from a lack of organizational understanding of what AI can realistically do, but it may also reflect a deeper systems issue: fragmented software ecosystems that cannot communicate internally or interface with AI models via APIs.
What is particularly concerning is that the dominant POS provider in the restaurant industry does not appear to offer this capability, while smaller platforms (e.g., Toast) do. This is notable given that the incumbent solution (Micros, under Oracle) is owned by a company that positions itself as an “AI leader.” If accurate, this suggests that even firms branding themselves as AI-centric may be failing to deliver meaningful AI functionality to their largest customers—and have been for years.
Hypothesis:
Assume, for the sake of argument, that the primary constraint on deploying accurate AI-driven solutions is not model sophistication, but rather the inability of a company’s internal software systems to communicate with each other or with AI tools via APIs.
This raises a broader question: What are the odds that future AI solutions will rely on data from a single, isolated software platform? If this fast-food example is representative, those odds appear extremely low.
Proposed Valuation Angle:
This leads to a potentially simple—but powerful—framework for assessing “AI readiness”:
What percentage of a company’s software platforms can actively communicate with one another and/or external AI tools via APIs?
The challenge, of course, is determining how much of this information is publicly observable or inferable by investors.
Additional considerations might include:
- Whether the firm has actual AI expertise at the C-suite level, rather than AI responsibilities being nominally assigned to a CTO whose role is primarily defensive (e.g., cybersecurity).
- The political capital of technical leadership within the organization—i.e., whether AI advocates can meaningfully influence operations-oriented executives.
- Governance structure and incentives that determine whether AI initiatives are integrated into core workflows or treated as peripheral experiments.
Observed Reality (Outside of Quant and Select Tech Firms):
In practice, most AI deployment I see in the U.S. economy falls into two categories:
- Replacing human customer support with AI-driven Q&A funnels.
- Reducing headcount and broadly instructing remaining staff to “figure it out with AI.”
Neither approach suggests deep operational integration or durable competitive advantage.
Questions for the Forum:
- Have you observed materially different AI adoption patterns in other sectors?
- What additional variables should be considered when designing an “AI readiness” valuation model?
- Should such a model be binary (e.g., AI-ready vs. not, based on a threshold like 75% system interoperability), or continuous (a weighted percentage-based score)?
- Are there existing disclosures, proxies, or diligence techniques that could make this framework investable?
Interested to hear others’ thoughts and experiences.
OpenAI added:
"Addendum: Additional Observations and Potential Investment Implications
A few additional thoughts that may help translate this concept from an abstract framework into something more actionable for valuation and investing:
1. AI Readiness Is More About Data Plumbing Than Algorithms
The marginal value of more sophisticated models appears far lower than the marginal value of clean, accessible, interoperable data at least in the short term. In many cases, even basic regression or rules-based models would materially improve outcomes if they had access to the full operational dataset. This suggests that AI readiness is less about “cutting-edge AI” and more about whether a firm has already done the unglamorous work of modernizing data architecture.
2. Legacy IT as a Hidden Liability
From a valuation perspective, legacy software stacks may represent an underappreciated form of technical debt. Two companies with similar margins and growth profiles may have radically different AI optionality depending on how fragmented their internal systems are. In this sense, poor system interoperability could function as a negative intangible asset—one that increases future capex, slows execution, and limits the ROI of AI initiatives.
3. Potential Public Proxies for AI Readiness
While direct API-level interoperability is rarely disclosed, there may be imperfect but useful proxies investors can track:
- Ratio of IT spend to revenue, adjusted for maintenance vs. growth capex.
- Disclosure of cloud migration progress (especially multi-cloud or API-first architectures).
- Frequency of core system replacements vs. bolt-on solutions in earnings calls.
- Depth of technical leadership beyond a single CTO (e.g., Chief Data Officer, Head of AI/ML).
- Language in 10-K risk factors that frames AI as operationally embedded vs. aspirational.
4. Organizational Friction as the Binding Constraint
Even where technical capability exists, organizational resistance often prevents deployment. Firms with highly centralized operations, strong process discipline, and measurable KPIs may be structurally better positioned to benefit from AI than more decentralized or politically entrenched organizations. This suggests that AI readiness is as much a management quality and governance issue as it is a technical one.
5. Market Mispricing Risk
Current market narratives often assume AI benefits will diffuse broadly and rapidly across incumbents. If AI adoption instead proves bottlenecked by software fragmentation and organizational inertia, then:
- Near-term AI gains may accrue disproportionately to firms already operating modern, integrated stacks.
- Incumbents with heavy legacy infrastructure may face longer and costlier transition periods than markets currently discount.
- Some “AI leaders” may be overstating near-term monetization potential, creating valuation risk.
6. Binary vs. Continuous Scoring
A continuous scoring model likely better reflects reality. AI readiness may exhibit nonlinear payoff dynamics—where moving from 30% to 60% interoperability matters far more than moving from 80% to 90%. This creates potential inflection points that could be useful for both fundamental and factor-based strategies."
One potentially important input when valuing companies on the basis of “AI readiness” is forming a view on where AI capabilities are ultimately likely to plateau. Having at least a rough upper bound on what AI can realistically do allows you to estimate the total addressable value of AI-driven productivity or margin expansion, against which a firm’s readiness score can be applied.
At a high level, the possible outcomes span a wide range. Conceptually, they might be framed as follows, from lower to higher impact:
AI improvement slows materially, and even the most advanced systems function primarily as very sophisticated search, summarization, and pattern-recognition tools. Gains are incremental and largely confined to cost reduction and workflow efficiency.
AI becomes capable of applying essentially all existing human knowledge to novel inputs, including generating combinations and permutations that have not previously been attempted. While it does not independently create fundamentally new knowledge, it dramatically accelerates problem-solving, optimization, and decision-making across domains.
AI reaches or exceeds human-level general intelligence in most domains, but remains constrained by architectural or alignment limitations (e.g., susceptibility to flawed inputs, adversarial prompts, or poorly defined objectives, aka "gullible"). Capabilities are transformative, but deployment still requires human oversight.
AI operates with minimal constraints and near-complete autonomy across cognitive tasks. While this scenario is often discussed in existential terms, from a valuation perspective it likely represents diminishing marginal relevance—at that point, traditional firm-level competitive analysis may no longer be meaningful.
Relevance to Valuation:
For investment purposes, scenarios (1) through (3) are likely the most actionable. Importantly, even under relatively conservative assumptions, the economic value unlocked by AI appears, in the near term at least, to depend far more on organizational readiness and system integration than on marginal improvements in model intelligence. In the long term, model intelligence importance dramatically increases in value.
Related Conceptual Note (Separate from the Valuation Model):
One hypothesis I’ve been considering is that a key structural limitation of AI may be the absence of innate “instincts” analogous to those found in biological systems. In nature, instincts function as foundational logic that governs prioritization, filtering, and learning. Human intelligence is not just accumulated knowledge, but knowledge organized around these core heuristics.
AI systems, by contrast, appear to rely almost entirely on externally defined objectives and training data. If this remains true, increasingly capable AI may still require significant guardrails and supervision, reinforcing the idea that organizational design and governance will remain critical determinants of economic value creation.
That said, this is speculative and separate from the core readiness framework—but it reinforces the broader point: technology alone is unlikely to be the binding constraint on AI-driven returns.
Interesting: Below are the original thoughts I fed to OpenAI, note how it removed any mention of AI being a potential threat...
"One potentially important factor in valuating companies based on AI is factoring in where AI technology will end up abilities wise. That way you have a rough total valuation or value of what AI can do for companies to apply a readiness score against.
Some options from “worst case scenario” to “best case scenario”:
1) AI doesn’t improve much more and the most advanced models are essentially just really good search engines.
2) Near AGI - AI gets good enough that it can effectively apply all current knowledge to any input from users, including combinations and permutations of knowledge that have never been tried or tested before. It can’t truly invent something new though.
3) AGI but gullible – full ability to surpass human intelligence while having an understanding of all knowledge, but since it doesn’t posses instincts, it can still be tricked.
4) Full AGI – It will likely kill us all since it would also posses the same core instincts shared across nature.
Fyi, this is sort of a new theory I’ve been thinking about when trying to conceptualize the current state and limitations of AI. Instincts in nature are essentially core logic that creatures are born with, and all learning throughout the beings life is filtered through and ordered/organized around those instincts. My theory (that is separate from the valuation model) is that since AI doesn’t have instincts built in, it will be very hard to keep if from constantly being tricked etc. I could very well be wrong on this front though.
"
OpenAI added:
"Addendum: Why AI Outcome Uncertainty Strengthens (Not Weakens) a Readiness-Based Framework
One implication of the wide dispersion in potential AI outcomes is that AI readiness should be viewed as an option-like attribute, rather than a deterministic earnings driver. When future capability distributions are highly uncertain, firms with flexible, interoperable systems and strong technical governance effectively own a call option on AI progress, while firms with fragmented or legacy stacks own something closer to a liability.
1. Optionality Over Precision
From a valuation standpoint, investors do not need to predict the exact endpoint of AI capabilities to price readiness. Even modest AI improvements (Scenario 1 or 2) can generate material economic value if a firm can deploy them quickly and broadly. Conversely, firms that lack integration may fail to monetize AI even under more aggressive capability assumptions.
2. Asymmetric Payoffs
AI readiness likely exhibits asymmetric return profiles:
This asymmetry suggests readiness should command a valuation premium, particularly in industries with high operating leverage or complex coordination problems.
3. Capital Allocation as a Signal
One underappreciated signal is how companies fund AI-related initiatives. Firms that treat interoperability and data infrastructure as long-lived strategic investments (rather than expensed experiments) may be better positioned than those pursuing fragmented pilot programs. Tracking capitalization policies, multi-year IT roadmaps, and integration-focused M&A could therefore improve model robustness.
4. Time-to-Value as the Critical Variable
In practice, the biggest differentiator may not be ultimate AI capability, but time-to-deployment. Companies that can translate incremental AI improvements into production systems within weeks or months may capture disproportionate value relative to slower-moving peers, even if all firms eventually reach similar technical endpoints.
5. Implication for Relative Valuation
If AI readiness is primarily about optionality and speed, it may justify:
In this sense, AI readiness could function less as a thematic bet and more as a cross-sectional differentiator within traditional fundamental analysis."
The full first draft valuation model open for feedback (again with help from OpenAI):
Proposed AI Readiness Scoring Rubric (0–100)
1. Systems Interoperability & Data Accessibility (30 points)
The single most important factor.
What to assess
Scoring guide
2. Data Quality, Ownership & Governance (20 points)
What to assess
Scoring guide
3. Organizational Capability & Incentives (20 points)
What to assess
Scoring guide
4. Deployment Speed & Execution Track Record (15 points)
What to assess
Scoring guide
5. Capital Allocation & Strategic Commitment (10 points)
What to assess
Scoring guide
6. Cultural & Operational Fit (5 points)
What to assess
Scoring guide
Interpreting the Score
Why This Works for Valuation
Highlights hidden liabilities (legacy IT, org friction) markets often ignore
In general, are most financial models additive, or do they use gating multipliers? Back in my computer science days, models that use gating multipliers were usually much more effective, but it definitely varies case by case. See:
"
Revised AI Readiness Rubric with Gating Factors
Core Insight
Gate A: Systems Interoperability (Required)
Without interoperable systems, AI cannot access the full decision context. No plumbing = no deployment.
Gate B: Data Integrity & Ownership (Required)
AI trained on untrusted data produces outputs that cannot be operationalized, regardless of model quality.
Gate C: Decision Authority & Incentives (Required)
AI that cannot act is a reporting tool, not a productivity driver.
Final Score=Base Score×A×B×C\text{Final Score} = \text{Base Score} \times A \times B \times CFinal Score=Base Score×A×B×C
You are correct, one signifigant refinement to the AI-readiness framework is that it shouldn’t be purely additive. In practice, AI value creation is conditional: certain failures don’t just reduce returns, they prevent deployment entirely. To reflect this, the model uses gating multipliers on a base score. If a firm lacks system interoperability (no real APIs), usable and governed data, or the organizational authority to act on AI outputs, its effective AI-readiness score collapses—potentially to zero—regardless of spend or pilot activity. This “multiply-by-zero” feature prevents false positives, explains why many AI narratives fail to translate into ROI, and better matches how AI actually scales in production environments.
"
Quae aspernatur est ut labore. Nobis ad aut sed. Illum aut voluptatem sint ipsum cum. Doloremque autem reiciendis ut quas qui aspernatur praesentium. Tempora autem nisi qui suscipit earum.
Vel eos adipisci tenetur explicabo qui voluptatibus. Ipsum ducimus odit nemo soluta.
Necessitatibus rerum numquam consectetur accusamus et. Aut exercitationem dolorum quidem ea et aperiam. Molestiae eveniet facere consectetur veritatis omnis exercitationem. Est et quae ad et eum nam velit ducimus. Quibusdam a facilis iusto voluptate velit. Dolores non sit quis optio beatae culpa architecto voluptatem. Tempora ut id ut eum.
Qui eius in omnis commodi aut ut velit. Vero et et ea soluta et dolorem possimus.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...