THE AI INTELLIGENCE DEFICIT The Missing Layer Between AI Investment and AI Value A Straven & Co Diagnostic Report January 2026 DOI: 10.5281/zenodo.18499586 SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6448880 License: CC BY 4.0 Straven & Co Inc. https://stravenandco.com --- 1. THE DEFICIT DEFINED The AI market has a missing layer. Technology is not the constraint. Artificial intelligence capabilities have never been stronger. Capital remains plentiful, with global AI investments totalling $202.3 billion in 2025, representing a 75% increase from the prior year. Ambition remains evident as organisations across various industries actively pursue AI transformation. In 2025, artificial intelligence captured nearly 50% of all global venture funding, up from 34% in 2024. What is lacking is intelligence. The market needs independent expertise to distinguish authentic AI value from marketing claims, verify capabilities beyond inflated assertions, and inform investment decisions that prevent costly failures. MIT's State of AI in Business 2025 report found that 95% of enterprise AI pilot projects do not achieve a measurable return on investment. We term this structural challenge the AI Intelligence Deficit. This deficit explains why 42% of organisations discontinued the majority of their AI initiatives in 2025, up from 17% the year before. It also explains why sophisticated investors continue to lose billions on initiatives that, in retrospect, were never properly understood. The Evidence The gap between AI investment and AI value is widening. Enterprise AI Pilot Success Rate: 5% deliver measurable P&L impact (MIT 2025) Initiative Abandonment Rate: 42% of companies in 2025, up from 17% in 2024 (S&P Global) Proof-of-Concept Failure: 46% of POCs scrapped before production (S&P Global) Overall AI Initiative Failure: 70-85% fail to meet expected outcomes (RAND Corporation) The Failures That Illustrate the Deficit Builder.ai collapsed into insolvency in May 2025, with its $1.5 billion valuation in ruins. An internal audit revealed that the company had reported $220 million in 2024 revenue when actual revenue was approximately $55 million. Microsoft, SoftBank, and Qatar's sovereign wealth fund were among the most prominent investors. The company operated without a Chief Financial Officer for 18 months while raising hundreds of millions in capital. Internal communications directed staff to minimise visibility of human labour in work marketed as AI-automated. By October 2025, the FBI had subpoenaed the former CFO to appear before a Manhattan grand jury, with prosecutors investigating potential wire fraud, securities fraud, and conspiracy. nate, an AI shopping application, raised venture capital on claims that the platform was fully automated by artificial intelligence, with human involvement limited to edge cases. The actual automation rate was zero percent. In April 2025, the Department of Justice charged the CEO with wire fraud, marking one of the first federal criminal prosecutions for AI-related misrepresentation. These failures were preventable. The expertise required to evaluate these claims exists within the market. The structural challenge is that this expertise operates outside the normal information channels that reach decision-makers, creating a gap between available intelligence and its application at the point of decision. --- 2. THE INFORMATION ARCHITECTURE OF FAILURE The AI Intelligence Deficit is not a mystery. It is the predictable result of how information flows in the AI market. Every individual involved in decision-making operates within an information environment. The quality of their decisions cannot exceed the quality of the information they receive. The artificial intelligence market exhibits a fundamentally unbalanced information landscape, not due to any conspiracy, but as a result of misaligned incentives and an uneven distribution of expertise. The sources that produce the most content have the least independence. The sources with the most expertise tend to generate the least amount of content. Decision-makers are flooded with the former and starved of the latter. Information volume is inversely correlated with information reliability. The voices that reach decision-makers most frequently are the voices least equipped, or least incentivised, to provide independent assessment. The AI Information Ecosystem Map We mapped the AI information ecosystem across two dimensions. Commercial incentive forms the vertical axis. Technical expertise forms the horizontal axis. Four distinct quadrants emerged, each exhibiting characteristic signal quality. Vendors possess technical expertise; however, their incentives may lead them to oversell. Amplifiers frequently emphasise engagement, resulting in exaggerated statements rather than rigorous analysis. Translators rely on Vendors and Amplifiers for information, which can serve to magnify errors rather than rectify them. Validators, who are independent experts, generate only limited content due to the extensive time required for comprehensive analysis, and because few individuals are able to undertake it. Consequently, the initial three groups provide decision-makers with substantial amounts of information, whereas Validators offer comparatively less. Much of the available information is lacking in depth and insight. The four quadrants: Vendors (High Expertise, High Commercial Incentive): AI providers and implementation partners. Signal quality: medium-low. Paid when solutions are sold, not when solutions succeed. Amplifiers (Low Expertise, High Commercial Incentive): Influencers and social media. Signal quality: low. Paid for engagement, not for accuracy. Translators (Low Expertise, Low Commercial Incentive): Journalists and analysts. Signal quality: medium-low. Rely on Vendors and Amplifiers for input; may magnify errors. Validators (High Expertise, Low Commercial Incentive): Independent experts with technical depth. Signal quality: high. Paid for accuracy, independence, and rigour. The Signal Degradation Curve The deficit compounds as information travels. We analysed the progression of AI capability assertions as they evolve throughout the ecosystem. At origin: Model achieved 94% accuracy on benchmark X under conditions Y with limitations Z. At the Vendor stage: Our AI achieves 94% accuracy, transforming how businesses operate. At the Media stage: New AI system outperforms humans with 94% accuracy rate. At the Amplifier stage: AI is now 94% as good as humans. Your job is at risk. At the Cultural stage: AI can do almost anything now. We need to adopt immediately or be left behind. The impressive number survives. The meaningful context disappears. By the final stage, the original finding is unrecognisable. The 94% that sounds impressive survives. The context that gives it meaning (benchmark, conditions, limitations) does not. By the time AI capability claims reach strategic planning conversations, the information has often degraded beyond utility. --- 3. THE MISSING LAYER The AI Intelligence Deficit is the absence of independent, technically sophisticated expertise at the point of decision. This layer is present. There are experts who can evaluate AI architecture, identify inflated claims, assess implementation risk, and distinguish genuine capability from demonstration theatre. But this expertise is not embedded in normal advisory channels. It is not present in vendor relationships. It is not accessible through media coverage or industry conferences. It exists, but it exists outside the information architecture that shapes most AI decisions. Why Internal Teams Cannot Fill the Gap Most organisations lack the necessary expertise in artificial intelligence to conduct a rigorous independent assessment. IT departments possess an in-depth understanding of infrastructure operations. Data teams possess extensive expertise in analytics. Neither of them necessarily understands the precise mechanisms through which AI may fail, the distinction between benchmark and real-world performance, or the patterns that differentiate genuine capability from advanced staging. The people making AI investment decisions are often structurally separated from the expertise required to evaluate AI claims. This is not a criticism. It is a description of how organisations are built. Leaders who possess expertise in strategy rather than AI architecture are responsible for making strategic decisions. The CDO Insights 2025 survey conducted by Informatica confirms this, with 43% of organisations identifying a deficiency in technical maturity as a significant obstacle to AI success, and 35% citing a skills shortage as another primary barrier. Only 26% of organisations can transition from proof-of-concept to full-scale production. Why External Channels Do Not Provide It External sources that influence decision-makers operate within incentive frameworks that reduce their independence. Vendors: Paid when solutions are sold, not when solutions succeed. Implementation Partners: Paid when projects proceed, not when projects are right-sized. Analysts: Paid for coverage and access, not for accuracy. Media: Paid for attention, not for verification. None of these relationships offers the independent judgement necessary for sound AI decision-making. What Builder.ai and nate Reveal Both failures were failures of intelligence. The missing layer was absent in both cases. Builder.ai's warning signs were visible. There was no CFO. Revenue claims were inconsistent with operational scale. Internal communications minimised human labour. These signals would have been legacy to independent technical experts. They were not legible to investors relying on normal due diligence channels. Following the collapse, the median revenue multiples for AI companies decreased from 35x to 25.8x, and monthly AI funding declined by 35% from its peak in March 2025. nate's claims of fully automated AI, despite zero actual automation, would not have withstood technical scrutiny. The scrutiny did not take place. The intelligence was present. It was not accessed. As technical evaluation becomes more accessible and fraud is increasingly documented, it is difficult to justify the absence of independent expertise. The key consideration is not whether businesses should seek external guidance before major AI decisions, but whether failing to do so is a responsible use of capital and organisational resources. --- 4. RESTORING INTELLIGENCE TO AI DECISIONS Organisations that consistently realise value from AI investments share a common characteristic. They have closed the AI Intelligence Deficit for themselves. They have access to independent expertise that can evaluate claims, assess architecture, and identify risk before capital is committed. This is not a competitive advantage. It is a baseline requirement for sound decision-making in the current environment. Three Disciplines for Closing the Deficit 1. Source Awareness Be mindful of the source of the information. Map sources to the AI Information Ecosystem. Weight information by independence and expertise, not by volume or reach. Information provided by Validators should be regarded as more significant than that from Vendors, Amplifiers, or Translators, even if it occurs less frequently. Practical Application: Before any major AI decision, identify the primary information sources that shaped the opportunity. If all sources have commercial relationships with the outcome, the intelligence is compromised. 2. Signal Tracing Determine the origin of the claims. When encountering AI capability claims, ask what the original finding was. Ask what the conditions and limitations were. Ask how many layers of translation have occurred. Practical Application: For any impressive-sounding statistic, identify the original source. If the source cannot be identified, the claim cannot be evaluated. 3. Intelligence Access Establish relationships with independent expertise before decisions are required. This may be internal capability development, external advisory relationships, or both. The key criterion is independence from commercial outcomes. Practical Application: Before committing capital to AI initiatives, engage expertise that is not paid by vendors, not motivated by project continuation, and technically capable of evaluating claims. The Straven Perspective At Straven & Co, we developed our approach in healthcare, an environment where the AI Intelligence Deficit has life-or-death consequences. A hallucination in a clinical context is not a technical glitch. It is a potential disaster. Diagnosis must precede design. Design must precede deployment. Most organisations neglect at least one of these stages. The majority of AI projects are unsuccessful due to precisely that reason. MIT's research indicates that acquiring AI tools from specialised vendors and establishing partnerships are successful approximately 67% of the time, whereas internal development achieves success only about one-third as often. The distinction lies in the existence or lack of independent expertise within the process. The AI Intelligence Deficit is structural, but it is not permanent. Organisations that invest in closing it, that establish access to independent expertise, that develop source awareness, that trace claims to origins, will realise the value that AI genuinely offers. The competitive advantage belongs to those who restore intelligence to AI decisions while others continue to operate in its absence. --- 5. CONCLUSION The AI market does not lack technology. It does not lack capital. It does not lack ambition. It lacks intelligence. It lacks the independent, technically sophisticated expertise required to evaluate claims, assess risk, and distinguish genuine value from sophisticated marketing. This deficit explains why 95% of AI pilots fail despite unprecedented investment. It explains why sophisticated investors lose billions on initiatives that, in retrospect, were never properly understood. It explains why the gap between AI investment and AI value continues to widen. The cases of Builder.ai and nate are not anomalies. They are symptoms of a market where the intelligence required for sound decisions is structurally absent from the channels that reach decision-makers. Closing the AI Intelligence Deficit is not optional for organisations serious about realising AI value. It demands source awareness, signal tracing, and access to expertise that is truly independent of commercial interests. The technology is ready. The capital is available. The question is whether the intelligence will be present where decisions are made. --- REFERENCES 1. Crunchbase, "6 Charts That Show The Big AI Funding Trends Of 2025," December 2025. 2. Crunchbase, ibid. AI captured close to 50% of all global funding in 2025, up from 34% in 2024. 3. MIT NANDA Initiative, "The GenAI Divide: State of AI in Business 2025." Based on 150 interviews, survey of 350 employees, and analysis of 300 public AI deployments. 4. S&P Global Market Intelligence, "2025 Enterprise AI Survey," surveying over 1,000 enterprises across North America and Europe. 5. Bloomberg, Financial Times, TechSpot reporting on Builder.ai collapse, May-June 2025. 6. Bloomberg, "US Subpoenas Builder.ai's Ex-CFO for Auditor Communications," October 20, 2025. Subpoena issued by U.S. Attorney's Office, Southern District of New York. 7. U.S. Department of Justice, Southern District of New York, United States v. Saniger, April 2025. CEO charged with wire fraud for misrepresenting AI capabilities. 8. Informatica, "CDO Insights 2025 Survey." Data quality and readiness (43%) and lack of technical maturity (43%) identified as top obstacles. 9. S&P Global Market Intelligence, ibid. 10. The Analytical Investor, "The $1.5B AI Fraud That's Reshaping Silicon Valley," June 2025. Post-collapse market analysis. 11. MIT NANDA Initiative, ibid. --- Published by Straven & Co Inc. https://stravenandco.com DOI: 10.5281/zenodo.18499586 License: Creative Commons Attribution 4.0 International (CC BY 4.0)