Will AI agents replace SaaS tools? An honest founder take

Is AI killing SaaS, and will AI agents replace SaaS tools? Partly, and the part they replace is narrower than the headlines suggest. Agents are absorbing the thin glue layer of software: point tools, single-purpose automations, and light CRUD apps that exist mainly to move a record from A to B. Systems of record such as CRMs, ERPs, and ledgers are holding up, and in 2026 most agents work on top of that existing SaaS through APIs.

The stock market tested the "SaaS is dead" story this year. The S&P 500 software and services index fell more than 26% between late January and an April low, in a selloff that sharpened after Anthropic released plug-ins for its Claude Cowork agent in early February 2026. By 6 October it had climbed to its highest level since November 2025, according to Reuters, as reported by the Economic Times. Investors priced in wholesale replacement, then took all of it back.

I run Gravity, an AI agent platform, so I have an obvious bias toward "agents win," and I will argue against my own interest where the evidence demands it. Below I separate the durable claim from the hype, show how the big vendors actually price agents as of October 2026, and give you a timeline you can plan against. For the constructive side of the shift, see how AI agents for SaaS pay off inside a software business today.

Key takeaways

• AI agents are absorbing the thin, workflow-glue layer of SaaS: point tools, single-purpose automations, and light CRUD apps. That part of the market is at real risk.
• Systems of record and deep domain platforms endure, and Gartner still forecasts software spending growth of 15.5% for 2026.
• The 2026 "SaaSpocalypse" fully reversed: software stocks fell more than 26% to an April low, then hit a 2026 high on 6 October.
• In 2026 agents sit on top of SaaS through APIs. Your CRM stays; the agent reads and writes to it for you.
• Pricing breaks first, and it is breaking into a hybrid. Salesforce, Fin, ServiceNow, and Microsoft pair a meter for agent work with seats or commitments; none has gone purely per use.
• Coding agents are a second threat: 32% of respondents in McKinsey's 2026 survey said their organization skipped at least one software purchase because it built the functionality in-house.
• Gartner expects over 40% of agentic AI projects to be canceled by the end of 2027, so plan for a multi-year shift.

The short answer

AI agents replace the workflow-and-interface layer of SaaS and lean harder on the data layer underneath. Value that lived in workflow plumbing migrates to a reasoning layer; value that lives in data, compliance, and deep domain logic stays put.

Think of any SaaS product as two things stacked together. Underneath sits a system of record: a structured database with hard-won rules about how your customers, invoices, or inventory behave. On top sits a system of engagement: the forms, buttons, dashboards, and "if this then that" automations that let a human drive the database. Agents are very good at the top layer and weak at safely owning the bottom one, so the top layer gets eaten first.

Bain's Technology Report 2025 draws a similar picture, describing agentic AI "rebundling control on a three-layer stack (bottom to top): systems of record, agent operating systems, and outcome interfaces." In my two-layer version, Bain's top two layers are what replace the old system of engagement. Its summary is the most balanced line I have read on the topic: "Disruption is mandatory, but obsolescence is optional."

If you want to put agents to work inside your own SaaS, the 8 best AI agents for SaaS companies compares what is available today. If your product is mostly the top layer, a glorified form over a table, you should be nervous; if it is the table itself, the record other systems trust, you have more runway than the headlines suggest.

What the discourse actually says

The discourse splits into two camps: executives and investors who say agents will absorb business applications, and operators who say ripping out a system of record is a multi-year, career-risking project. Each is right about a different layer, and the most-quoted bullish line says more than most people repeat.

Nadella's quote, read to the end

The most-cited version of the bullish case comes from Satya Nadella, in a December 2024 episode of the BG2 podcast with Bill Gurley and Brad Gerstner. A clip circulated again in May 2025, when Windows Central published a transcript, which is why the quote is sometimes dated to then. He was describing Microsoft's own approach to Dynamics:

"the notion that business applications exist, that's probably where they'll all collapse, right in the agent era, because if you think about it, right, they are essentially CRUD databases with a bunch of business logic. The business logic is all going to these agents, and these agents are going to be multi repo CRUD, right? So they're not going to discriminate between what the back end is. They're going to update multiple databases, and all the logic will be in the AI tier, so to speak. And once the AI tier becomes the place where all the logic is, then people will start replacing the back ends, right?"

That last sentence matters. An earlier version of this post cut the quote before it and read Nadella as conceding that the databases survive. He says the opposite: once the logic lives in the AI tier, people replace the back ends too. Nadella goes further than I would, and on the data layer I disagree with him. I expect the system of record to be the last layer to go, for three practical reasons:

• Something has to be canonical. An agent that updates "multiple databases" still needs one of them to be the source of truth, along with the audit trail and access controls that come with it.
• Migration is slow and risky. Few teams volunteer to move years of customer, finance, or clinical records early.
• Compliance attaches to the record. Auditors ask where the data lives and who changed it, and the answer has to be deterministic.

Microsoft's own pricing points to a slower path. Covering the April 2026 earnings call, CRN reported that Microsoft detailed a business model "evolving beyond per-seat applications and consumption-based cloud infrastructure to an AI-era model combining the two," and that it now has more than 20 million paid Microsoft 365 Copilot seats. The CEO who forecast the collapse of business applications is, for now, selling agents by the seat and by the meter.

The investor and operator view

Andreessen Horowitz (a16z) made a quieter and, to me, sharper version of the argument in its December 2024 enterprise newsletter: "Per-seat is no longer the atomic unit of software." When AI does the work, the value metric shifts from how many humans log in to how much output gets delivered. I think that is the most durable part of the thesis, and I return to it in the pricing section.

Operators push back on two fronts. An agent that occasionally gets a database write wrong will not be allowed near production finance data. And Anthony Pierri's May 2025 LinkedIn post walks through everyday tasks, such as scanning Slack messages, glancing at a CRM pipeline view, or moving sticky notes in Figma, where a screen beats a chat box. The mistake on both sides is treating "SaaS" as one thing that either dies or survives.

What the 2026 SaaSpocalypse actually showed

The 2026 selloff showed that investors were ready to price in the wholesale replacement of software, and that the businesses underneath kept growing anyway. Software stocks slid from late January to an April low, then recovered to 2026 highs by early October as earnings forecasts rose. That arc fits the narrower glue-layer story far better than "SaaS is dead."

The trigger and the slide

The slide that earned the "SaaSpocalypse" nickname sharpened in the first week of February 2026, after Anthropic released plug-ins for its Claude Cowork agent. CNBC described the tools as "designed to handle complex professional workflows that many software and data providers sell as core products," and Reuters named the launch of Anthropic's legal plug-in as one of the triggers. The figures quoted for it cover different windows, so here is the arc by date and outlet:

The recovery, and the skeptics

By autumn the index had won back everything it lost in the selloff. Projected 2026 earnings growth for the software sector rose to 20.6%, from 13.8% at the end of March, according to LSEG data cited by Reuters, as reported by the Economic Times. Valoir CEO Rebecca Wettemann put it plainly: "The whole SaaSpocalypse didn't happen anywhere near as fast as some of the people on Wall Street thought it would."

A few voices said as much during the selloff itself. Nvidia's Jensen Huang dismissed the idea that "the software industry is in decline and will be replaced by AI" as "the most illogical thing in the world." Futurum's Rolf Bulk took the middle line: "There's likely to be cannibalization of SaaS by AI-driven workflows and that will impact the multiple the sector trades on."

My read on what it means

The market priced in the right mechanism at the wrong speed. According to Reuters, as reported by the Economic Times, investors feared that companies could use AI to build applications in-house more cheaply and so buy less from traditional software providers. What arrived in 2026 was a slower and more selective version: some purchases skipped in favour of home-built tools, which I cover in the build-vs-buy section below, plus agents taking over glue work on top of existing systems and steady pressure on seat-based pricing. The risk has not passed for everyone, though. Zacks' Brian Mulberry, as paraphrased by the Economic Times, thinks the sector could face its biggest test in the second half of 2027, when more data-centre capacity may make AI coding tools a bigger threat to traditional software companies.

What gets disrupted vs what survives

Thin workflow tools get disrupted; systems of record, deep domain platforms, and compliance surfaces survive. That is my scorecard from building agents that have to work with real customer stacks, and Forrester draws the line in the same place: "Horizontal point-solution SaaS vendors will be challenged," while "Vertical- or domain-specific SaaS vendors will have a greater chance of survival."

What gets disrupted, and fast

• Point tools and single-purpose automations. Anything whose whole value is "connect app A to app B and copy a field" is squarely in the blast radius. This is the same territory I covered in AI agents vs workflow automation, where rigid trigger-action chains lose to a model that can reason about the goal.
• Light CRUD apps. The internal tool that is a form over a spreadsheet, the lightweight tracker, the bespoke admin panel. An agent plus a database can replace a lot of these, and increasingly does.
• Thin reporting and "glue" layers. Dashboards nobody opens, status-update tooling, the connective tissue between bigger systems. Connective tissue is exactly what agents are best at.

One more group is exposed, on a slower clock: legacy tools that agents cannot operate. SaaStr's Jason Lemkin makes this point in a section titled "Old SaaS Can't Even Be Operated by an Agent." A product with no usable API gets routed around, which is a slower kind of decline.

What survives, and arguably gets stronger

• Systems of record. Your CRM, ERP, ledger, EHR, and data warehouse. These hold the canonical truth other systems depend on. Agents need them more, because an agent without a reliable source of truth is just confident guessing.
• Deep domain platforms. Software encoding years of regulatory, accounting, or clinical logic does not get reconstructed by a prompt. The moat is the thousand edge cases handled correctly under the screen.
• Compliance, audit, and security surfaces. The parts buyers cannot legally hand to a probabilistic system without controls. These harden into requirements that agents must satisfy, which I come back to in the timeline below.

The evidence that the core is still growing

If AI were killing SaaS wholesale, spending and revenue at the core would be shrinking. As of October 2026 they are still growing:

• Software spending. Gartner's July 2026 forecast puts worldwide software spending at $1,468 billion for 2026, up 15.5% from $1,271 billion in 2025. That segment includes AI software, so read it as software overall.
• A major system of record. Salesforce reported $11.3 billion in revenue for its fiscal 2027 second quarter (released 26 August 2026), up 11% year over year, including $456 million from Informatica. Agentforce ARR passed $1.5 billion, though Salesforce widened that metric this quarter to include Slackbot and Headless 360, so it does not compare cleanly with earlier quarters.
• SaaS spending. Forrester projects global SaaS spending rising from $318 billion in 2025 to $512 billion in 2028 and $576 billion in 2029, which it reads as evidence "that the enterprise core isn't vanishing, even as it transforms."
• Workflow platforms. ServiceNow raised its full-year subscription revenue forecast in April 2026, and CEO Bill McDermott predicted AI product sales close to $1.5 billion for 2026, a forecast Fortune put at 50% above his prior one. His summary: "There are two worlds that are integrating. They are not taking each other out. They are cooperating." Investors still drove the shares down as much as 14% in after-hours trading, which shows how far sentiment had drifted from the numbers.

The uncomfortable middle is the mid-market SaaS suite that is mostly workflow with a shallow data model. Those vendors face a real choice: become the system of record their customers trust, or become a feature an agent calls. Many will not make that jump.

Agents sit on top of SaaS, for now

In 2026 agents mostly work through existing SaaS. They log in to Salesforce, read the pipeline, draft the follow-ups, update the records, and hand a summary back. The SaaS tool is still there; the agent is a new layer of intent above it, talking to it through APIs and increasingly through shared protocols.

The incumbents now say so themselves. In Salesforce's August 2026 results, Marc Benioff described opening up "the data, workflows, business logic, actions, and governance inside Salesforce and making it available to every agent, model, and interface." Business logic is Nadella's phrase from 2024. My reading is that Salesforce's answer to agents taking over the logic is to expose that logic to outside agents and keep the record.

Analysts expect this layer to spread quickly. In August 2025 Gartner predicted that 40% of enterprise applications would be integrated with task-specific AI agents by the end of 2026, up from less than 5% at the time, and estimated that by 2028 "a third of user experiences will shift from native applications to agentic front ends." That is the interface layer moving, which is the part I expect to go first.

APIs give agents permissioned access to systems of record without reimplementing twenty years of domain logic, and interoperability standards, which I track in our coverage of AI agent future trends for 2026, give them an auditable way in. That is why the most useful early agents look like better front doors to existing tools; compare the best Zapier alternatives using AI and our head-to-head on Gravity versus Zapier. The agent absorbs the glue; the apps it touches keep doing what they do well.

Where the screen still wins

Pierri's objection deserves a straight answer. Some tasks stay visual: glancing at a pipeline board, scanning a busy channel, arranging ideas on a whiteboard. A chat box is a worse tool for those, and I do not expect agents to take them over soon. What the agent front end takes over first is multi-step, cross-app work, the jobs where a person clicks through four tools to finish one task. Gartner's "a third of user experiences" by 2028 still leaves most experiences on native screens, which sounds about right to me.

Describe the outcome, not the workflow

Agents eat the glue layer first because they change what users have to specify. Traditional SaaS makes you translate your goal into the software's vocabulary: pick this menu, configure that rule, map these fields, build the seven-step automation. You do the translation work, and the tool executes the steps you designed.

Agents flip that. You state the outcome, "reconcile last month's invoices against the bank feed and flag anything over a thousand dollars that doesn't match," and the agent works out the steps. I have written about this at length in my piece on outcome-first software, and it is why the workflow layer goes first. Once a machine can plan the steps, a product whose entire job was letting humans hand-build those steps loses most of its value.

A workflow tool executes a fixed graph you authored. An agent reasons about a goal, chooses tools, handles the branches you did not anticipate, and recovers from failures. That capability gap is why I am bullish on the category while staying skeptical of the "SaaS is dead" headline. The systems underneath still have to be right, though: reconciling invoices needs a ledger and a bank feed that are correct and trusted. That division of labor is the realistic shape of the next few years, and it is what the examples in the top AI agent use cases for H1 2026 demonstrate.

The other threat: building instead of buying

Coding agents threaten SaaS by a different route: they let customers build some functionality themselves and skip the purchase. McKinsey's State of AI Global Survey 2026 found that "nearly one-third of respondents (32 percent) report that their organizations have decided against purchasing at least one software product or feature because they were able to build the functionality in-house using agentic coding tools."

That is the strongest real-data version of the "AI is killing SaaS" fear I have seen, and it is the fear investors traded on in February. Forrester's list of investor worries includes the idea that "vibe coding will allow startups to replicate the features of complex SaaS platforms, eroding their moat."

It hits the same glue layer from the other side. Agents on top of SaaS hollow out the workflow tools you already pay for; coding agents stop you buying the next one. Neither route does much to the system of record. A team that builds its own approval tracker in an afternoon still keeps its ledger, its CRM, and its identity provider, and the new tool reads from them. If you sell a point product, both pressures land on you at once. If you own the record, home-built tools mostly become new callers of your API.

The pricing shift: from seats to seats plus meters

Pricing is the part of SaaS that cracks first, and as of October 2026 it has cracked into a hybrid. Salesforce, ServiceNow, and Microsoft sell seats for humans alongside a meter for agent work, and Fin pairs its per-outcome agent with a per-seat helpdesk. None of these incumbents has moved to pure per-use pricing; even Fin's seat-free option, for teams on another helpdesk, comes with minimum commitments.

The logic behind the shift is simple. Per-seat pricing assumes a human sits in a chair and uses the software. One agent can do the work of several seats around the clock without logging in as ten users, so the seat count stops tracking the value delivered. The a16z newsletter's December 2024 example was customer support: companies paid Zendesk "per support agent ($115/month/seat)," but once AI can handle ticket resolution, "the natural pricing metric becomes successful outcomes." Zendesk had already announced outcome-based pricing for its AI agents in August 2024, and its own blog now puts the price at $1.50 per automated resolution.

How incumbents price AI agents, as of October 2026

The figures below come from each vendor's own pricing page, read on 12 October 2026, unless the source column says otherwise. All are in USD. These pages change often (Salesforce's says its pricing is "subject to change"), so check before you budget.

Two details stand out. First, the hybrid shows up even at Fin, which prices its agent per outcome but still charges per seat for the humans on its own helpdesk. Second, consumption units carry fine print. Salesforce's page states that "Unused Flex Credits do not roll over into subsequent subscription terms," so prepaid capacity you have not used by the end of a term is gone.

What Salesforce's 2026 deals signal

Salesforce, long the reference point for per-seat SaaS, spent 2026 buying the machinery of usage-based pricing:

• Fin, formerly Intercom. Salesforce signed a definitive agreement on 15 June 2026 to acquire Fin for approximately $3.6 billion and completed the deal on 10 September 2026, adding a customer base of more than 30,000 companies. Any comparison that pits Intercom against Salesforce as rivals is now out of date.
• m3ter. A metering and rating platform "purpose-built for consumption-based monetization," acquired in a deal completed on 1 July 2026.

Analysts expect the contract shift to continue. Forrester writes that "SaaS vendor contracts are primarily based on seats. This will shift to consumption (and even outcome-based) pricing as AI agents are deployed." Bain's advice to incumbents is to "own the data, lead on standards, and price for outcomes, not log-ons." For how agent platforms themselves changed their prices this year, see AI agent pricing trends in 2026.

At Gravity I chose flat monthly plans with no per-seat fees, because a seat count says little about how much work an agent does. Free is $0 with no card, Mini Pilot is $15 a month, Auto Pilot is $20 a month, and you can buy more usage from your account if you run out. Gravity is in public beta.

A realistic timeline

Glue-layer disruption is under way now, mid-market workflow suites face pressure toward 2030, and systems of record persist beyond that as services agents call. Predictions are cheap, so here is the data I anchor that to.

Adoption is broad and getting deeper, while value lags. McKinsey's State of AI: Global Survey 2026, published 25 August 2026 and fielded from 4 May to 8 June with 1,719 participants in 97 countries, reports that:

• Nearly nine in ten respondents report regular use of AI in at least one business function.
• 44% say AI is scaling across their enterprise, up from 38% a year earlier.
• 40% of respondents from organizations with more than $1 billion in annual revenue report scaling AI agents, up from 27%, while smaller organizations stayed flat at 22%.
• 37% report that AI has contributed positively to EBIT, essentially unchanged from 2025.

Wide adoption with flat profit impact is what I would expect during a glue-layer transition, where agents save time on small tasks long before they change how a business runs.

The agent-specific outlook is more sobering. Gartner forecasts that over 40% of agentic AI projects will be canceled by the end of 2027, due to escalating costs, unclear business value, or inadequate risk controls. The same June 2025 release predicts that at least 15% of day-to-day work decisions will be made autonomously through agentic AI by 2028, up from 0% in 2024. Both numbers argue for precision about timing: meaningful change, and a long way from "SaaS is gone."

So here is my timeline, hedged on purpose:

• Now through 2027: glue-layer disruption is under way. Point tools, light automations, and thin CRUD apps lose ground; agents run on top of systems of record via APIs; incumbents add meters next to seats; coding agents replace some purchases outright. Many ambitious agent projects also get canceled, for the cost, value, and risk-control reasons Gartner names.
• 2027 to roughly 2030: mid-market workflow suites face their reckoning. Those that became systems of record survive; those that stayed thin get absorbed into agent platforms. Outcome-based meters become normal.
• Beyond 2030: deep domain platforms and systems of record persist, increasingly as services agents call. Gartner's best-case projection has agentic AI driving about 30% of enterprise application software revenue by 2035, surpassing $450 billion, up from 2% in 2025. The interface layer is what collapses; the data layer stays.

The single biggest variable, in my view, is reliability. Until an agent finishes the job correctly the overwhelming majority of the time, with audit trails a compliance team accepts, the replacement story stays capped at low-stakes glue work. That is why I keep telling people the near-term winner is "agents plus your existing SaaS."

Frequently Asked Questions

Is AI killing SaaS?

AI is killing one layer of SaaS: thin point tools, single-purpose automations, and light CRUD apps that agents can now replace. Core platforms are still growing; Salesforce reported revenue up 11% year over year in August 2026. The market's early-2026 panic also fully reversed: the S&P 500 software and services index fell more than 26% from late January to an April low, then reached its highest level since November 2025 on 6 October.

What was the SaaSpocalypse?

It is the nickname for the 2026 software stock selloff, which sharpened in the first week of February after Anthropic released plug-ins for its Claude Cowork agent, including a legal plug-in. Forrester estimated that over $1 trillion of software market capitalization was erased in seven days, and the S&P 500 software and services index kept sliding until an April low. By early October the index had recovered to a 2026 high as earnings forecasts rose.

Will AI agents replace SaaS tools completely?

No, not completely or soon. Agents are absorbing the workflow-glue layer: point tools, single-purpose automations, and light CRUD apps. Systems of record like CRMs, ERPs, and ledgers endure because they hold the canonical data and deep domain logic that agents depend on.

Are AI agents replacing SaaS tools right now in 2026?

Partly. In 2026 most agents run on top of existing SaaS through APIs, reading and writing records on your behalf while the underlying software stays in place. Thin automations and basic internal tools are being displaced, and in McKinsey's 2026 survey 32% of respondents said their organization skipped at least one software purchase because it built the functionality with agentic coding tools.

Why do people say SaaS will collapse in the agent era?

The most-cited version comes from Satya Nadella, who argued in a December 2024 episode of the BG2 podcast that business applications are essentially CRUD databases with business logic, and that the logic is moving to agents. He went further, predicting that once the logic lives in the AI tier, "people will start replacing the back ends." I disagree on that last step: I expect systems of record to be the last layer to go, because they hold the canonical data, audit trails, and compliance obligations.

What kind of SaaS is most at risk from AI agents?

Thin, workflow-heavy products with shallow data models are most exposed: point tools that copy fields between apps, single-purpose automations, light CRUD apps, and basic dashboards. If a product's entire value is letting humans hand-build steps a machine can now plan, it is in the blast radius. Forrester expects horizontal point-solution vendors to be challenged, while vertical or domain-specific vendors have a greater chance of survival.

How are SaaS companies pricing AI agents in 2026?

Mostly with hybrids that pair seats or commitments with a meter for agent work. As of October 2026, Salesforce Agentforce lists $2 per conversation or Flex Credits at $500 per 100,000 (about $0.10 per standard action), alongside per-user add-ons at $125 a month. Fin, formerly Intercom and owned by Salesforce since 10 September 2026, charges $0.99 per outcome, and Zendesk's blog puts its price at $1.50 per automated resolution.

How does agent pricing differ from per-seat SaaS pricing?

Per-seat pricing charges per human user, which breaks down when one agent does the work of many seats. Vendors are adding a meter tied to work, such as actions, conversations, or outcomes, usually alongside seats. Gravity takes a simpler route: flat monthly plans with no per-seat fees, where Free is $0 with no card, Mini Pilot is $15 a month, Auto Pilot is $20 a month, and you can buy more usage if you run out.

What is the realistic timeline for agents disrupting SaaS?

Glue-layer disruption is already under way and runs through 2027, though Gartner expects over 40% of agentic AI projects to be canceled by the end of 2027 due to escalating costs, unclear business value, or inadequate risk controls. Mid-market workflow suites face pressure toward 2030, while deep platforms and systems of record persist as services agents call.

The bottom line

AI is not killing SaaS; it is splitting it. Agents replace the workflow-and-interface layer of software while depending more heavily on the data-and-domain layer underneath. The glue gets eaten, the systems of record get stronger, and for the next several years the dominant pattern is agents sitting on top of your existing stack.

If you are a buyer, run this checklist before your next renewal:

• Which tools exist only so humans can click? Those are the first candidates for an agent, or for something your team builds in-house.
• Which tool owns the canonical record? Protect it, and make sure agents can reach it with proper permissions.
• Does the vendor expose APIs or agent access? A tool an agent cannot operate will get routed around.
• Is the renewal priced on seats, meters, or both? Model what happens to the bill if an agent takes over much of the seat-holders' routine work.
• What are the rules on consumption units? Check rollover and minimums; Agentforce's unused Flex Credits, for example, do not roll over into the next subscription term.

Don't rip out your CRM to chase a headline. If you build SaaS, the survival move is to become the trusted record that agents call. And if you want to feel the outcome-first shift yourself, that is what Gravity is built around: state the result you want and an expert-built agent runs the task in about 60 seconds. Gravity is in public beta.

Sources

• BG2 podcast (YouTube): Satya Nadella, BG2 w/ Bill Gurley & Brad Gerstner (published 12 December 2024)
• Windows Central, Kevin Okemwa: transcript of Nadella's BG2 remarks on agents and business applications (published 9 May 2025)
• CRN: Microsoft Q3 earnings, Nadella says AI agents change how customers pay (published 30 April 2026)
• Andreessen Horowitz: AI Is Driving A Shift Towards Outcome-Based Pricing (published 19 December 2024)
• Anthony Pierri on LinkedIn: How will AI agents replace SaaS? (posted 15 May 2025)
• Bain & Company: Will Agentic AI Disrupt SaaS? Technology Report 2025 (published 23 September 2025)
• CNBC: software stocks sell off after Anthropic's new AI tools (published 6 February 2026)
• Reuters via Yahoo Finance: global software stocks hit by Anthropic wake-up call on AI disruption (published 4 February 2026)
• Forrester, Kate Leggett: SaaS as we know it is dead, how to survive the SaaS-pocalypse (published 11 February 2026)
• Economic Times (Reuters report): US software shares scale fresh 2026 highs as AI disruption worries fade (published 6 October 2026)
• SaaStr, Jason Lemkin: AI Isn't Killing SaaS. SaaS Is Killing Itself. (accessed 12 October 2026)
• Gartner: worldwide IT spending forecast to grow 14.2% in 2026 (published 27 July 2026)
• Salesforce: second quarter fiscal 2027 results (published 26 August 2026)
• Fortune: ServiceNow earnings, forecast and AI product sales (published 23 April 2026)
• Gartner: 40% of enterprise apps will feature task-specific AI agents by 2026 (published 26 August 2025)
• McKinsey: The State of AI: Global Survey 2026 (published 25 August 2026)
• Gartner: Over 40% of Agentic AI Projects Will Be Canceled by End of 2027 (published 25 June 2025)
• Salesforce: Agentforce pricing (accessed 12 October 2026)
• Fin: pricing (accessed 12 October 2026)
• Zendesk newsroom: first in CX industry to offer outcome-based pricing for AI agents (published 28 August 2024)
• Zendesk blog: outcome-based pricing (published 16 April 2026, updated 13 May 2026)
• Salesforce: signs definitive agreement to acquire Fin (published 15 June 2026)
• Salesforce: completes acquisition of Fin (published 10 September 2026)
• Salesforce: agreement to acquire m3ter, with completion note dated 1 July 2026 (published 8 June 2026)