SaaS vs AI Agents: Is SaaS Losing Its Edge?
For fifteen years the formula was simple: rent software by the seat, pay every month, and let the vendor handle hosting and upgrades. The SaaS vs AI agents debate puts that formula in question. If an AI agent can read your inbox, update your CRM, raise a purchase order and reconcile an invoice, why pay for ten human seats in each of those tools? Investors asked exactly that in early 2026, and the answer moved markets. This article looks past the headlines and asks a practical question: what should a business that depends on SaaS actually do about it?
What Happened: The 2026 'SaaSpocalypse'
In early February 2026 a sharp sell-off hit software stocks. Forrester, writing on 11 February 2026, put the loss at more than one trillion dollars of software market value. Commentators quickly named it the 'SaaSpocalypse'. The trigger was not a single product launch but a growing belief that AI agents, software that can plan and carry out multi-step work across applications, would reduce the number of people who need to log into business tools.
It is worth being precise about what the sell-off measured. It measured investor expectations about future revenue, not a sudden collapse in how companies use software. Your ERP did not stop working that week. But expectations matter, because they push vendors to change pricing, packaging and product strategy, and those changes land on your invoices.
Why Per-Seat Pricing Is Under Pressure
Most SaaS contracts charge per user. That model works when value grows with headcount: more salespeople, more CRM seats. Agents break the link between value and headcount in three ways:
- Fewer humans in the tool: if an agent handles first-line support tickets, the support desk may need fewer seats even though it processes more tickets.
- Work moves above the application: an agent can orchestrate several systems through their APIs. The user's daily experience happens in the agent interface, not in each vendor's screens.
- Value becomes measurable per task: when software completes a task end to end, buyers start asking to pay per resolved ticket, per processed invoice or per qualified lead rather than per login.
Forrester expects seat-based contracts to move toward consumption and even outcome-based pricing, and notes that some vendors already sell flexible credits usable across human seats and agents. For buyers, that means the next renewal may look very different from the last one.
SaaS vs AI Agents: What Actually Changes
The most useful way to frame SaaS vs AI agents is not 'one replaces the other'. Agents need something to act on. They read and write data in systems of record: the ERP that holds stock and costs, the CRM that holds customers, the accounting system that holds the ledger. What changes is which layer captures value.
- Systems of record stay: data, permissions, audit trails and transactions still need a reliable home. These are hard to replace and hold their value.
- Thin workflow tools are exposed: products that mainly provide a screen on top of a simple process, with low switching costs, are most at risk because an agent can perform the same steps.
- Vertical, data-rich software holds up better: Forrester argues that domain-specific vendors with proprietary data and complex industry needs are better placed than generic tools.
- The orchestration layer is new territory: whoever controls the agent layer that sits above several applications gains influence over how work is done.
What Is Not Changing
Several things that made SaaS attractive remain true. Someone still has to host, secure, back up and update software. Compliance requirements, from e-invoicing rules to data protection, still need to be implemented and maintained. Agents also introduce new needs of their own: permission models that limit what an agent may do, logs that show what it did and why, and human approval for risky actions such as payments or price changes.
Reliability is the other constant. A human user notices when a screen looks wrong. An agent may confidently repeat a mistake across hundreds of records. Businesses will keep paying for software that gives them control, traceability and predictable behaviour, whatever the pricing model.
What This Means for Your Software Budget
For an operations or finance leader, the debate becomes concrete at renewal time. A few practical steps help:
- Map your stack by role: list each tool and mark it as a system of record, a workflow tool or a reporting tool. Workflow tools are where agents will change usage first.
- Watch seat utilisation: if agents or automations reduce active users, make sure contracts let you reduce seats rather than paying for empty licences.
- Model new pricing before signing: consumption and outcome pricing can be cheaper or more expensive than seats. Estimate volumes, set caps and ask how overages are billed.
- Check API access and terms: an agent is only as useful as the APIs it can call. Confirm rate limits, costs and whether the vendor allows third-party agents on your data.
- Keep ownership of your data: make sure you can export data in usable formats, so you are not locked into one vendor's agent strategy.
- Consolidate where it makes sense: overlapping tools that each charge per seat are the first candidates to merge or replace.
Where Custom Software Fits In
When agents shift value toward data and orchestration, companies with unusual processes face a choice. They can wait for each vendor to add agent features that roughly fit, or they can own the pieces that make them different. Custom software is not a replacement for every SaaS tool. Accounting, email and HR software are usually best bought. But the layer that connects your ERP, your warehouse, your dealer network and your production floor, and the agent workflows that run on top of it, is often where a tailored build pays off.
A sensible pattern for 2027 looks like this: keep proven SaaS and ERP products as systems of record, expose their data through clean integrations, and build custom workflows and agent tasks around the processes that are specific to your business. That keeps you flexible as pricing models shift, because the logic that matters lives in software you control.
A Sector View: Manufacturing and Wholesale
In manufacturing, distribution and wholesale, the shift looks less dramatic than in sales or marketing software. Much of the value sits in systems of record: item masters, bills of materials, stock by location, customer-specific prices and open orders. Agents will not replace those. Where they will help is in the work around them: reading supplier confirmations and updating expected delivery dates, answering dealer questions about stock and order status, flagging late shipments, matching invoices to goods receipts and preparing replenishment suggestions for a planner to approve.
For these companies, the practical question is whether their data is clean and reachable. An agent cannot answer a dealer's stock question if stock lives in three spreadsheets and an ERP that has no usable API. Investing in integrations, consistent master data and clear approval rules is the groundwork that makes any agent, bought or built, useful.
Risks to Plan For
- Cost unpredictability: usage-based agent pricing can spike when volumes grow or when an automation loops. Budgets need monitoring and limits.
- Security and permissions: an agent with broad access to several systems is a high-value target. Apply least-privilege access and review actions it can take.
- Regulation: in the EU, the AI Act's transparency duties largely apply from 2 August 2026, while high-risk obligations were postponed by the Digital Omnibus to 2 December 2027 for stand-alone systems. Check whether any of your use cases fall into those categories.
- Vendor churn: a market under pressure means consolidation. Some tools you rely on may be acquired, repriced or discontinued.
Frequently Asked Questions
Are AI agents going to replace SaaS?
Not wholesale. In the SaaS vs AI agents debate, the likely outcome is that agents change how SaaS is used and priced. Systems of record such as ERP, CRM and accounting remain, while simple workflow tools with low switching costs face the most pressure.
What is the SaaSpocalypse?
It is the nickname for the sharp sell-off in software stocks in early 2026, driven by fears that AI agents would reduce the number of paid software seats. Forrester estimated more than one trillion dollars of software market value was lost.
Will per-seat pricing disappear?
It is unlikely to vanish quickly, but analysts expect a shift toward consumption-based and outcome-based pricing, often mixed with seats. Expect hybrid contracts with credits that can be used by humans and agents.
Is usage-based pricing cheaper for businesses?
It depends on volumes. It can be cheaper when usage is low or seasonal and more expensive when usage grows. Model expected volumes, ask for caps and monitor consumption.
Should we cancel SaaS tools and build our own?
Rarely as a blanket decision. Commodity functions are usually best bought. Custom software makes sense for processes that differentiate your business or for integrations that tie several systems together.
What should we check before letting an AI agent access our systems?
Check API access and terms, limit permissions to what the task needs, log every action, require human approval for financial or irreversible actions, and confirm how data is stored and processed.
Does the EU AI Act affect AI agents used in business?
It can. Transparency obligations largely apply from 2 August 2026, and high-risk obligations for stand-alone systems now apply from 2 December 2027. Whether you are affected depends on the use case, so seek legal advice for specific cases.
Final Thoughts
The SaaS model is not dying, but its pricing and its centre of gravity are moving. The winners will be businesses that know which systems hold their data, which workflows set them apart and how much they are really paying per outcome. Codefacture builds custom software for manufacturing, supply chain, automotive, wholesale and packaging companies in the UK, Australia, New Zealand, the Gulf and Türkiye. If you want to connect your existing systems and build agent-ready workflows around them, we can help you plan the next step.