Von SaaS vs. McKinsey zu "McTool"

Sep 10, 2025

How Agentic AI and Forward-Deployed Engineers Are Disrupting Classic Software Go-to-Market

Companies Want Solutions, Not Tools

Companies have little interest in buying yet another isolated software tool—they demand solutions to concrete business problems. Classic Software-as-a-Service (SaaS) products, however useful individually, hit their limits when they don't seamlessly deliver results. A prominent tech investor compared enterprise customers to grandparents with new tech: "Enterprises buying AI are like your grandma getting an iPhone: they want to use it, but they need you to set it up." In other words: The value should be tangible without the company having to handle complex setup itself. The reality is: Software that doesn't directly improve processes or is cumbersome to implement won't succeed long-term.

This development is particularly visible in the AI world. AI-powered systems (agentic AI—AI agents that autonomously execute actions) only unlock their value when embedded into a company's individual workflows. A generic AI demo impresses short-term; without specific adaptation to operational needs, "the software doesn't stick"—it doesn't become a permanent part of daily work. Companies don't want a toolkit they must assemble themselves, but a key that opens the lock of their problem.

From SaaS Tool and McKinsey to "McTool": Product and Service Converge

For a long time, consulting and implementation services were separate from go-to-market. Investors avoided consulting revenue like the plague. However, the crises triggered by generative AI at McKinsey & Co. point the way forward: Consulting without context and implementation is unnecessary—ChatGPT handles that. At the same time, nobody has been waiting for the next SaaS tool that an Account Executive drops off at the company gate with a Post-it note saying "Good luck with implementation." Both models are faltering.

Instead of "McKinsey vs. SaaS tool," the success formula is now "McTool"—the fusion of McKinsey-style, context-rich problem-solving and an agentic AI stack. Many AI products are delivered as products but often miss their purpose without intensive customization. Winners will be those who deliver a complete, customer-specific tailored package—consulting know-how directly embedded in the product.

A pioneer is Palantir—let's call it the Palantir model. Palantir deploys Forward-Deployed Engineers (FDEs) directly into customer projects to configure their proprietary platforms precisely. Result: Instead of generic software, the customer receives an immediately deployable solution for their specific use case. This "solution-first" approach was long considered difficult to scale, but Palantir has built a remarkable business on it. With the new AI portfolio (Palantir AIP), growth is accelerating: In the most recent quarter, Palantir's commercial US revenue grew 64% year-over-year—impressive evidence of strong enterprise demand for highly customized AI solutions. Palantir's market cap now ranks among the top 25 in global tech—the market significance of this model is unmistakable.

Forward-Deployed Engineers are the key to this shift. They bridge product and customer-specific solution: on-site (or closely) with the customer, they translate business problems into robust configurations. Palantir popularized the model, but it's finding imitators: Even AI platform providers like OpenAI or Anthropic are hiring FDEs to anchor their solutions in large enterprises. The principle: Hands-on support as part of the product. This costs more initially (personnel, budget) but creates a moat—the deeply integrated solution isn't easily replaced. You're trading margin for moat upfront for long-term market power.

The McTool paradigm turns the old SaaS game upside down. In the 2010s, Product-Led Growth (PLG) was the holy grail—Atlassian, Slack, Dropbox showed how far self-explanatory tools could carry. In the AI era: Complex problems require integrated solutions. Salesforce, ServiceNow, and Workday already proved in the cloud shift that intensive implementation pays off. Today, with AI, this necessity is greater—and AI paradoxically helps make integration faster. The trend is clear: Forward-deployed teams and products are growing together. Whoever masters both from a single source—consulting plus tool—will define tomorrow's market leaders.

Usage Beats License: The End of the Classic ARR Model

The shift also affects pricing models. The classic SaaS subscription (ARR per user) is under pressure. Usage-based pricing (consumption-based) is increasingly taking hold. AI services especially drive this: Costs (e.g., compute power) and value often correlate with actual usage—consumption-based prices seem more sensible and fair.

The numbers support the trend: Already 85% of software firms have introduced or are planning usage-based pricing models; among the largest providers, 77% use at least partial consumption-based pricing. The shift is accelerating through AI: Nearly half of companies with usage models introduced them only in the last two years—largely in response to AI innovations. Marc Benioff, CEO of Salesforce, puts it plainly: "We have seat-based products for humans, and we have consumption products—those are for agents and robots." As long as humans were the main users, seat-based subscriptions dominated. With scaling AI agents, consumption-based tariffs are needed.

Microsoft and Adobe are already reacting appropriately with pricing adjustments for AI features (e.g., prices per 1,000 AI requests). In parallel, AI-as-a-Service offerings are emerging where licenses aren't central, but the AI performance delivered. Revenue is generated depending on intensity and utility of the solution. For providers, this means more volatile revenues and farewell to beloved fixed ARR—but also greater upside when utility scales.

Palantir as Proof-of-Concept: Solution-First Scales

Counterargument: "That doesn't scale!" Palantir shows it scales—if you set it up right. The company has sold solution-first for years: first solve the problem, then broaden. Each new use case is initially accompanied by intensive engineering; however, this creates features that are later productized and made accessible to other customers. Palantir itself describes: FDEs configure the platform for one customer, but some of the most valuable product features originated in the field and were then generalized.

Meanwhile, Palantir is growing not only in government but strongly commercially. Since launching the AI platform in 2023, metrics have shot up—the company exceeded its own projections and became profitable for the first time. Investors rewarded the course: In early 2025, the stock jumped over 20% in one day after strong results and optimistic AI outlook. The message: Deeply integrated, AI-powered solutions with embedded engineers are real growth drivers. Accordingly, many GenAI startups are orienting themselves similarly: Better a few large, service-heavy projects with measurable impact, then pour learnings into templates—and thus leverage scale effects.

Other major players are also testing solution-first. Salesforce is investing in "Einstein" copilots for industries and enterprises. Microsoft integrates copilots into the entire portfolio and monetizes them partly separately (e.g., Microsoft 365 Copilot). The extreme demand for GitHub Copilot (over 20 million users, including 1.3 million paying subscribers) shows: Companies pay extra for productivity-enhancing AI solutions.

The Endangered Middle: When Generic Is Too Little and Simple Is Too Much

The middle of the SaaS market is particularly under pressure—offerings that are neither ultra-light and cheap nor highly complex and mission-critical. Simple tools at the lower end (e.g., time tracking for a few euros per month) won't be replaced by expensive AI projects. At the upper end sit massive ERP/CRM systems, deeply woven into processes—these aren't replaced overnight.

In between, however, lies a broad field of mid-tier SaaS that offers "merely" better interfaces or moderate process improvements—often at considerable subscription fees. Precisely these applications are predestined to be substituted by AI agents plus integration teams. Why? Because here the cost-benefit calculation is worst: too generic for 100% problem-solving, too expensive as commodity. A capable AI team can relatively quickly configure an agent that handles the exact use case—and cleanly integrate it into the IT landscape via FDE. The customer receives a tailored solution instead of an off-the-shelf compromise. If costs (e.g., usage-based) align with added value, the old SaaS tool gets cancelled.

Existing SaaS providers face a choice: secure downward (become so cheap, simple, and indispensable that nobody bothers with individual AI) or develop upward—toward a solution-oriented platform that delivers services itself when needed. Otherwise, a SaaS twilight looms: Many mediocre products will disappear—displaced by integrated AI alternatives that better merge with customer processes.

Paradigm Shift: From Cloud to AI Agent—Like Nokia to iPhone

At its core, this is a paradigm shift in interaction. The jump from on-premise to cloud primarily changed deployment—not how we work with software. The interface remained forms, mouse, keyboard. Today's AI wave changes exactly that: We interact via generative AI and agentic systems.

The leap from classic tools to agentic AI resembles the switch from Nokia phones to iPhone. Nokia once dominated, but the iPhone changed usage (touch, apps, internet in hand) so radically that incumbents stumbled. Similarly, classic SaaS applications could fare poorly when ChatGPT-like interactions and automated agents become the norm.

Instead of opening windows in a dozen tools, manually transferring data and clicking buttons, we'll tell the assistant in future: "Create a forecast from this data and send the report to the team." The AI handles the rest—across formerly separate silos. This is the ChatGPT moment: Excel spreadsheet, data room, email archive are queried by voice—and deliver answers or completed tasks. This new interface layer sits above classic software and cannibalizes parts of it. Who needs specialized UIs when the agent uses APIs in the background?

Fundamental systems (databases, ERP core tables, etc.) remain—but as infrastructure in the background. Value creation shifts: Not the provider with the best UI wins, but the one who makes their functions best available via AI agents (or offers such agents themselves). Microsoft has recognized this and integrates Copilots into Office, Windows, GitHub & Co. to secure dominance. For new players, this is the chance to score with AI-first experiences.

Conclusion: The Game Is Being Reshuffled

Agentic AI + Forward-Deployed Engineers paint a future where classic SaaS and PLG lose radiance—not because software becomes unimportant, but because delivery and usage are changing. Solutions instead of tools is the motto: Whoever comprehensively solves an acute problem can count on revenue and loyalty; whoever delivers "just software" must hope customers bear implementation and integration burden. The next generation of successful B2B providers will work solution-first, bill usage instead of licenses, and deliver AI-powered interactions that seemed like science fiction recently. For the SaaS industry, a new era begins—the cards are being reshuffled.

www.scalehouse-capital.com

Quellen