External signals: SaaS and AI disruption

Raw source preview

Raw, unprocessed file text shown below. Preview truncated for readability.

---
title: "External Signal: SaaS AI Disruption — PE Theses and What They Mean for EmpathyIQ"
source: Video transcript + fact-checked against public sources
date: 2026-04-15
status: active
tags: [strategy, pricing, market-context, ai-disruption, external-signals]
---

# External Signal: SaaS AI Disruption — PE Theses and What They Mean for EmpathyIQ

## Summary

A video circulating in April 2026 summarises the AI disruption theses of three major software PE firms — Thoma Bravo, Vista Equity, and Coatue — and uses them as a frame for what is happening to public SaaS valuations. All five factual claims in the video have been independently verified against public sources. The reports referenced are partially or fully accessible (see below). This document evaluates what is relevant to EmpathyIQ's strategy and what is not.

---

## Fact-Check: Are the Claims True?

### Public market declines (Atlassian -63%, Asana -58%, Workday -47%)

**Verdict: Accurate — these are 2026 YTD figures as of approximately late March 2026.**

These figures are confirmed by stock data and press coverage from April 2026. They are not 2025 figures. The broader framing — that enterprise SaaS is under significant downward valuation pressure — is accurate and well-evidenced. For comparison, the BVP Nasdaq Emerging Cloud Index has declined substantially YTD 2026.

### Workday CTO left to become Member of Technical Staff at Anthropic

**Verdict: Accurate in substance; the "demotion" framing is editorial.**

Peter Bailis, Workday CTO since May 2025, left in March 2026 to join Anthropic as a Member of Technical Staff (MTS) — an individual contributor engineering role. He is working on reinforcement learning systems; his enterprise/HR background connects to Anthropic's push to build Claude-powered HR applications. The "demotion" characterisation is accurate colloquially (he gave up a C-suite title) but MTS at Anthropic is a standard designation for senior technical contributors, not a hierarchical step down.

### Thoma Bravo: $183B AUM, Miami LP meeting, mission-critical SaaS thesis

**Verdict: Fully accurate and unusually well-sourced — the LP slides were publicly released.**

$183B AUM confirmed. The Annual Meeting was held in Miami in March 2026. Holden Spaht (Managing Partner) publicly posted the LP slides to LinkedIn; they were covered by Augment Market, The Information, and others. The aircraft maintenance and hospital billing examples are directly from the slides. This is not a rumour or secondhand interpretation — it is Thoma Bravo's stated, on-record position.

> **The LP slides / source material:** Accessible via the Holden Spaht LinkedIn post and the Augment Market summary (augment.market/pulse/the-worlds-largest-software-buyer-says-the-market-is-wrong-about-ai).

### Vista Equity Partners: ~$100B AUM, ~90 portfolio companies, data sovereignty thesis

**Verdict: Accurate. AUM is slightly understated (actual: $107B+). Thesis is publicly documented.**

The data sovereignty framing is not from private LP materials — Robert Smith has stated this publicly in multiple interviews, including a Vista-published piece ("Software's Next Chapter", February 2026) and a Private Markets Insights interview (November 2025). The key quotes: *"CEOs haven't done a good job preserving sovereignty... They were on an ARR race... they've leaked intellectual property."* This is on-record commentary, not inference.

> **Source material:** vistaequitypartners.com/insights/softwares-next-chapter-robert-f-smith-on-the-next-era-of-enterprise-platforms/

### Coatue: ~$70B AUM, per-seat to per-output shift, $200B → $5.5T services-as-software

**Verdict: Fully accurate. The $5.5T figure comes from a specific, publicly available Coatue publication.**

Lucas Swisher (Coatue Management) published "AI Is Hitting a New Inflection Point" on March 26, 2026 (coatue.com/c/takes/ai-is-hitting-a-new-inflection-point). The $5.5T "services-as-software" figure, the per-seat to per-output framing, and the "selling work not software" thesis are all from this piece. Thomas Laffont (Coatue's technology lead) made related points publicly at the Upfront Summit in February 2026.

**Note:** The $5.5T figure is contested. HFS Research puts the same market at $1.5T. Coatue's figure treats the entire global services economy as the addressable market — it is the most expansive framing. This does not mean it is wrong, but the size should be treated as directional.

> **Source material:** coatue.com/c/takes/ai-is-hitting-a-new-inflection-point

---

## The Three Theses, Summarised

| Firm | Core Thesis | What They're Doing About It |
|---|---|---|
| **Thoma Bravo** | Mission-critical SaaS (aircraft maintenance, hospital billing) is AI-resistant. The market is wrong to sell it off. | Taking mission-critical SaaS private at depressed valuations — $12.3B Dayforce acquisition, $10.55B Boeing Aviation IT acquisition |
| **Vista Equity** | Incumbents can win IF they own their data. Companies that chased ARR gave up their moat and leaked IP through bad contracts. | Pushing portfolio companies to build around data sovereignty and restructure customer contracts |
| **Coatue** | The entire SaaS model is being rewritten. Value is shifting from per-seat access to per-output delivery. The addressable market grows 25x in this world. | Publishing the thesis publicly; investing in companies positioned to sell outcomes not access |

**No consensus.** Three of the world's most sophisticated software investors have materially different views on how AI disrupts their portfolios. This is important context: the outcome is genuinely uncertain.

---

## Strategic Relevance to EmpathyIQ

### What is directly relevant

#### 1. The Vista Equity thesis validates the knowledge layer as EmpathyIQ's core moat

Vista's argument is that SaaS companies that chased ARR by making everything easy to plug in also made it easy to rip out, because they never built proprietary data assets that customers couldn't recreate elsewhere. The companies that will win are the ones whose data — accumulated in a structured, proprietary form — becomes irreplaceable.

**This is exactly what EmpathyIQ's knowledge/compound layer is designed to do.** Every study run through the platform enriches a structured research repository that:

- Is tagged against ~15 metadata categories
- Integrates external research not built in the platform
- Becomes more valuable with every additional study
- Cannot be replicated by switching to a competitor — the history stays

The strategic implication is not to change course but to **be more explicit about the moat**. The knowledge layer is not a feature or a tier differentiator. It is the data sovereignty play. EmpathyIQ's switching cost story should be: *a customer who has run 50 studies through the platform has a structured, queryable research library that took years to build. They cannot take that to SurveyMonkey or Qualtrics.*

**Additional implication:** Vista's point about bad contracts leaking IP is relevant to how EmpathyIQ's own customer contracts are structured. Contracts should:
- Make clear that customer research data belongs to the customer
- Protect EmpathyIQ's proprietary methodology, confidence framework logic, and AI training data from flowing into customer-controlled LLMs
- Ensure that research conducted on external platforms and imported into EmpathyIQ does not create unclear data ownership

This is worth reviewing before enterprise contracts are signed.

#### 2. The Thoma Bravo thesis validates the enterprise infrastructure play — and raises the bar for what "mission-critical" means

Thoma Bravo's argument is that software becomes AI-resistant when removing it would create catastrophic organisational or regulatory risk. Aircraft maintenance records and hospital billing are their examples — not because these are flashy, but because the downside of failure is irreversible.

**EmpathyIQ is not yet mission-critical in that sense.** It is not there yet. But the trajectory — described in the pricing model and the in-depth strategy doc — is correct: if EmpathyIQ becomes the organisation's structured, validated system of record for customer understanding, and non-research teams start depending on it directly, then removing it means losing the institutional memory of customer knowledge. That approaches mission-critical.

The implication is that **the Level 3 enterprise play (Research Infrastructure, £15K–£40K/month) is the right destination**, and the path there should be prioritised. Point-solution sales (Level 0) are useful for adoption but do not build switching costs. The moat is built at Level 2 and above.

#### 3. The Coatue thesis on per-output pricing is a near-term strategic question, not a future one

Coatue's argument: the economic unit is shifting from a seat (a person who has access) to a unit of work (a completed study, a validated report, an answered question). In this world, the TAM is not the software budget but the services budget.

**EmpathyIQ's current pricing model is already moving in this direction.** The three-dimension structure (platform subscription + research activity + confidence layer) explicitly moves away from pure per-seat. The pricing doc states: *"Customers are not paying for seats or features. They are paying for the ability to run research properly, capture knowledge, and embed insights."*

However, the Coatue thesis taken to its conclusion would mean EmpathyIQ sells outcomes: a validated survey delivered to a defined audience, a research question answered against the knowledge base, an annual research programme. The current pricing model prices access to a system that produces these outcomes — not the outcomes themselves.

This is worth exploring as a packaging question, not necessarily a pricing architecture change. For example:
- "10 validated studies per year with full Trust Centre QA, delivered in [X] days" as a contract commitment
- Research programme packaging that bundles execution, confidence, and knowledge at a fixed annual commitment
- Outcome SLAs (e.g. guaranteed data quality threshold) that differentiate from tool-access competitors

This framing also has a positioning implication: EmpathyIQ should be able to say it sells *trusted research output*, not *research software*. That is a services-as-software positioning.

---

### What is not directly relevant

**The specific valuation declines of Atlassian, Asana, and Workday** are not relevant to EmpathyIQ. These are large public companies whose valuations reflect macro SaaS sentiment, not the research-tools competitive landscape specifically. EmpathyIQ competes in a narrower market and will not be priced on public SaaS multiples in the near term.

**The Workday CTO story** is colour commentary about the broader talent and confidence signals in enterprise software. It is not actionable for EmpathyIQ, though it reinforces the narrative that traditional enterprise software companies are struggling to retain technical leadership against AI-native organisations.

**The disagreement between the three PE firms** is informative as a reminder of genuine uncertainty but does not require EmpathyIQ to take a position on which firm is right. EmpathyIQ is early-stage and should build toward the outcome that any of these theses would reward: a product that is increasingly mission-critical, that builds a proprietary data moat, and that eventually sells research outcomes rather than research access.

---

## Summary: What to Do With This

| Thesis | Verdict for EmpathyIQ | Action |
|---|---|---|
| Vista: data sovereignty as moat | Directly validates the knowledge/compound layer strategy | Make this explicit in positioning and sales. Review enterprise contract language before first deal. |
| Thoma Bravo: mission-critical SaaS is AI-resistant | Validates the Level 3 enterprise infrastructure destination | Prioritise the path to Level 2/3. Point-solution sales don't build the moat. |
| Coatu