Divya van Mahajan

How AI Eats Consulting

ai-daily-brief-podcast

How AI Eats Consulting

Overview

This episode of The AI Daily Brief examines the accelerating convergence of AI product companies and consulting/professional services, using OpenAI’s expansion into consulting-style “forward-deployed engineer” services as the central case study. The broader argument is that AI is simultaneously a revenue driver for traditional consultants in the short term and an existential threat to their business model in the medium-to-long term. The host also covers Meta’s formal announcement of its Superintelligence Labs, Apple’s potential outsourcing of Siri, and Cursor’s mobile web app launch.

Source video: Not available (internal recording; no public URL provided)

Prerequisites

  • Basic understanding of how large language models (LLMs) and foundation models work
  • Familiarity with enterprise software business models (SaaS, product-led growth, systems integrators)
  • General awareness of major AI labs: OpenAI, Anthropic, Meta AI, Google DeepMind
  • Understanding of consulting industry structure (McKinsey, Accenture, PwC, GSIs/Global Systems Integrators)
  • Familiarity with Palantir’s business model and go-to-market strategy

Main Points

Meta Officially Launches Superintelligence Labs (MSL)

  • Mark Zuckerberg announced the formation of Meta Superintelligence Labs (MSL) in a staff memo, describing it as Meta’s central effort to develop superintelligence.
  • MSL consolidates foundations, product, and FAIR teams, plus a new lab for next-generation model development.
  • Alexander Wang (former Scale AI CEO) will lead MSL and serve as Meta’s Chief AI Officer; Nat Friedman (former GitHub CEO, active AI investor) also joins in a leadership role.
  • A list of 11 new hires was released, drawn heavily from OpenAI, with additional hires from Anthropic and Google DeepMind.
  • The Information expressed skepticism about the team’s cohesion, noting Wang has no foundation model track record and that the group risks “combustibility” due to large egos and intense pressure.
  • The host argues that simultaneous high-profile hiring can create its own momentum and that media skepticism may be underweighting the excitement of a shared mission.

Apple Reportedly Considering Outsourcing Siri to OpenAI or Anthropic

  • Bloomberg’s Mark Gurman reported Apple has held discussions with both OpenAI and Anthropic about using their models to power a future version of Siri.
  • Apple’s internal LLM Siri project continues, but executives under new project lead Mike Rockwell tested third-party models and reportedly found Anthropic’s Claude performed best.
  • Apple has approved a multi-billion dollar cloud budget but no final decision has been made; some Apple engineers have signaled they may leave for Meta or OpenAI.
  • Some commentators framed this as Apple abandoning its full-stack identity; the host disagrees, arguing Apple has no viable path to competing on AI talent and must partner or acquire rather than build alone.
  • The host contends average consumers care only that Siri works, not how it works, reducing brand risk from third-party models; and that Apple’s device distribution (billions of installed devices) remains valuable to model providers like Anthropic.

Cursor Launches Mobile-Accessible Web App for AI Coding Agents

  • Cursor released a web application allowing users to manage AI coding agents from a browser on desktop or mobile.
  • This follows a progression of interface expansions: background agents (May), Slack integration (June), and now the web app.
  • This is the first time Cursor has been accessible on mobile without a Slack workaround.
  • Early user response has been enthusiastic; the host frames this as a natural evolution toward voice and mobile-first interaction with coding agents.

OpenAI Moves into Consulting via Forward-Deployed Engineers

  • The Information reported OpenAI is building a forward-deployed engineer (FDE) team, hiring roughly a dozen engineers (several from Palantir) to embed directly with enterprise customers.
  • The service centers on fine-tuning — customizing OpenAI models (e.g., GPT-4o) on a customer’s proprietary data to solve company-specific problems.
  • Entry-level spend required is approximately $10 million, positioning this as an enterprise-only offering.
  • The FDE model was pioneered by Palantir; Palantir describes FDEs as engineers who embed with customers to configure existing platforms rapidly, distinguishing them from traditional consultants by speed and technical creativity rather than bespoke solution-building.
  • The host frames OpenAI’s move less as a radical pivot and more as a validation of FDEs as a necessary enterprise playbook component; OpenAI is also partnering with implementation firms (Tribe AI, Fractional) and major GSIs (PwC) rather than going it alone.

The “Palantirification” Trend Across the AI Industry

  • Andreessen Horowitz published a piece in June titled “Trading Margin for Moat: Why the Forward-Deployed Engineer is the Hottest Job in Startups”, formalizing the trend.
  • The a16z argument: during platform shifts (cloud → mobile → AI), companies that win are often those with complex, deeply integrated implementations rather than pure product-led growth (PLG) plays.
  • Historical examples: Salesforce, ServiceNow, and Workday all launched with below-average gross margins (54–63%) due to heavy implementation costs but achieved dominant market positions and combined valuations that dwarf top PLG companies.
  • The AI platform shift is distinct because AI itself can automate much of the integration and implementation work, accelerating deployment timelines and lowering the cost of building moats.
  • The host observes that owning the customer relationship — not just model quality — is increasingly OpenAI’s strategic priority.

AI’s Existential Pressure on Traditional Consulting

  • Consulting firms like Accenture, McKinsey, and PwC have seen strong short-term revenue growth from AI-related engagements, but structural pressure is building.
  • The Economist reported that Accenture’s market cap dropped ~$60 billion from its February 2025 peak, with new bookings declining for both one-off projects and managed services.
  • PwC’s Chief AI Officer acknowledged the firm has begun cutting prices because internal AI use is saving staff time, and clients are demanding they pass those efficiencies on.
  • The host’s own firm uses AI to conduct large-scale discovery work (interviewing hundreds to thousands of people and producing actionable agent use-case recommendations) in days and at less than one-tenth the cost of a traditional consulting equivalent — work that was previously impossible or required months.
  • The host argues AI is simultaneously eliminating the most commoditized consulting work (discovery, data gathering) while leaving the high-value advisory work; firms that adapt could build enduring positions, but the competitive field now includes software companies, AI labs, and neo-consultancies all offering technology and services simultaneously.

Key Concepts

  • Forward-Deployed Engineer (FDE): An engineer who embeds directly with a customer to configure and customize an existing software platform for that customer’s specific problems, combining speed and technical flexibility without building solutions from scratch.
  • Fine-tuning: The process of further training a pre-existing AI model on a specific dataset (e.g., a company’s proprietary data) to improve its performance on domain-specific tasks.
  • Palantirification: The trend of technology product companies adding embedded, services-heavy go-to-market motions — modeled on Palantir’s FDE approach — to deepen customer relationships and build switching-cost moats.
  • Product-Led Growth (PLG): A go-to-market strategy in which the product itself drives user acquisition and expansion, typically characterized by self-serve onboarding, no required sales process, and high gross margins.
  • Meta Superintelligence Labs (MSL): Meta’s newly announced AI research and product division consolidating all AI efforts under Alexander Wang, with a mandate covering both foundational research and product development.
  • Global Systems Integrators (GSIs): Large consulting and technology services firms (e.g., Accenture, PwC, Deloitte) that implement and integrate enterprise software across complex organizational environments.
  • Reinforcement Learning (RL): An AI training technique that rewards a model for achieving specified goals and penalizes it for undesirable behaviors, used to align model outputs with specific business metrics.
  • LLM Siri: Apple’s internal project to rebuild Siri on a large language model foundation, running in parallel with Apple’s reported external discussions with OpenAI and Anthropic.

Summary

The central argument of this episode is that the AI industry is undergoing a structural convergence in which product companies — most visibly OpenAI — are adopting consulting-style, services-heavy go-to-market models through forward-deployed engineers, while traditional consulting firms face simultaneous revenue pressure from AI-driven efficiency gains and new competitive entrants. The “Palantirification” of AI — embedding engineers directly inside enterprise customers to customize and operationalize AI platforms — is emerging as a dominant strategic pattern, validated by Palantir’s market success, endorsed by Andreessen Horowitz, and now being adopted by OpenAI. At the same time, incumbent consulting firms like Accenture are seeing market cap erosion and downward pricing pressure as AI reduces the cost of work they have historically charged premium fees for. The host’s broader message is that every layer of the enterprise technology stack — model providers, software platforms, implementation firms, and consultancies — is collapsing into a single competitive arena, and that the firms most likely to endure are those that move quickly to own deep customer relationships rather than clinging to legacy margin structures or full-stack independence.