---
title: "The 'AI replaces SaaS' story was wrong. SaaS metering AI is what's actually killing it."
description: "HubSpot, Adobe and Salesforce are now charging customers based on AI usage, according to The Information's late April 2026 reporting. The \"AI replaces SaaS\" prediction got the direction right and the mechanism wrong. SaaS isn't dying because AI competes with it. SaaS pricing is breaking because the usage clocking layer that vendors are bolting on is changing the deal. After 120+ AI projects across 15+ industries, I've seen this exact pattern hit three UK SMBs in the last quarter. One client went from £1,200 a month in SaaS AI add-ons to £180 a month on raw API. Here's the decision framework, with three buyer responses and one worked example."
canonical: https://richardbatt.com/blog/saas-metering-ai-killing-saas
date: 2026-05-05
author: Richard Batt
tags: [SaaS, AI Pricing, Vendor Strategy, SMB AI]
type: blog_post
---

# The 'AI replaces SaaS' story was wrong. SaaS metering AI is what's actually killing it.

_HubSpot, Adobe and Salesforce are now charging customers based on AI usage, according to The Information's late April 2026 reporting. The "AI replaces SaaS" prediction got the direction right and the mechanism wrong. SaaS isn't dying because AI competes with it. SaaS pricing is breaking because the usage clocking layer that vendors are bolting on is changing the deal. After 120+ AI projects across 15+ industries, I've seen this exact pattern hit three UK SMBs in the last quarter. One client went from £1,200 a month in SaaS AI add-ons to £180 a month on raw API. Here's the decision framework, with three buyer responses and one worked example._

**Richard Batt** — AI implementation specialist. 120+ projects across 15+ industries, serving SMBs (5-200 employees) worldwide from Middlesbrough, UK (working globally). Contact: richard@richardbatt.com · https://richardbatt.com

The story we kept being told over the last year was that AI agents would replace SaaS tools. Why pay HubSpot when an AI agent can run your CRM directly? Why pay Salesforce when a Claude agent can talk to your database? It made for confident LinkedIn posts. It hasn't happened, and won't, at the speed predicted. But something else has happened, and it's hitting UK SMB budgets right now.

According to The Information's reporting in late April 2026, also picked up by MarketScreener on April 27, HubSpot, Adobe and Salesforce have begun charging customers based on AI usage rather than seat count alone. The metering is happening inside products SMBs already buy, not as a separate AI product. So the "AI replaces SaaS" narrative got the outcome partly right but the mechanism completely wrong. SaaS isn't being replaced. SaaS pricing is being restructured around AI consumption, and the bill at the end of the quarter looks different.

After 120+ AI projects across 15+ industries, I've watched this same metering pattern hit three UK SMB clients in the last quarter. One went from £1,200 a month in SaaS AI add-ons to £180 a month on raw model APIs once we redesigned the workflow. Another stayed on the SaaS but cut feature usage by 40% to control the usage clock. The third decided to wait six months and re-test pricing. All three responses are valid. The decision depends on factors that are predictable if you know what to look for.

## What's actually changing

The shift is mechanical. SaaS vendors with large existing customer bases (HubSpot at over 200,000 customers, Adobe and Salesforce at corresponding scale) are bundling AI features into their products. The bundling started as "free included AI" two years ago, moved to "AI add-on for £20 per seat" last year, and is now moving to "AI usage credits, top up when you run out" in 2026.

The mechanism resembles mobile data plans. Customers get a base allocation each billing cycle and exceed it on busy months. From there the choices are simple: upgrade the plan, reduce usage, or pay over-limit fees. For UK SMBs running standardised marketing or sales workflows, the over-limit fees can show up as £400 to £1,500 a month surprises that nobody expected when the budget was set.

The vendor logic is straightforward. AI inference has a real per-call cost. Bundling it as an unlimited included feature would absorb that cost into shrinking margins. Metering AI consumption pushes the variable cost back to the customer, where it's easier to manage at the product economics level. This isn't malicious. It's how the business model has to work for the vendor to stay solvent.

But it's not how UK SMB budgets are typically structured. Most SMBs I work with budget SaaS as a fixed monthly line. Variable usage-based costs require a different finance posture and a different way of managing usage. Most haven't built either yet.

## Three buyer responses

When the usage clock shows up, SMBs have three honest options. I've now seen all three play out across clients in the last six months.

**Response 1: rip and replace.** Drop the SaaS AI feature entirely. Run the same workflow on raw model APIs (OpenAI, Anthropic, Google) connected via Zapier, Make, or n8n. This works when your usage volume is high enough that the SaaS metering is more expensive than the raw API plus integration cost, and when you have someone in-house who can manage the integration. The 30-person services firm I mentioned above moved from £1,200 a month in HubSpot AI add-ons (mostly automated email drafting and lead summarisation) to £180 a month on Anthropic's API plus a small Zapier setup. Net saving: £12,240 a year. Setup cost: about 14 hours of operations manager time.

**Response 2: ration and monitor.** Keep the SaaS but actively manage the usage clock. Set monthly caps. Pull weekly usage reports. Train the team on which features burn credits fastest. This works when the SaaS is genuinely embedded in the workflow and a rip-and-replace would cost more than the usage clocking does. A 40-person UK property firm I worked with took this route. Their HubSpot AI usage was driving £600 a month in metered fees. They didn't want to leave HubSpot but couldn't absorb the cost. We capped automated email drafts at 200 a week per user, redirected lead summarisation to an internal Claude prompt, and brought metered cost back to £180.

**Response 3: wait and re-test.** Do nothing for six months. Re-test pricing in October when vendor pricing has stabilised and at least three competitors have published comparable terms. This works when the usage clocked cost is currently low enough not to hurt and when your team doesn't have capacity to redesign workflows yet. A 12-person consultancy I spoke with in March picked this. Their AI add-on cost was £140 a month on Adobe Creative Cloud's AI features. They decided that was tolerable for now and they'd revisit at year-end.

The three responses aren't ranked. Each is right under different conditions. The mistake is to default to one without checking the conditions.

## A decision framework

Here's the test I now run with clients facing SaaS AI metering. Four questions, in order. The answer to each one narrows the response.

| Question | If yes | If no |
| --- | --- | --- |
| Is the usage clocked AI feature genuinely doing 5+ hours of work a week? | Continue to question 2 | Cancel the AI feature, save the cost, move on |
| Is the workflow stable enough to rebuild on raw APIs? | Consider rip-and-replace | Stay on SaaS, ration usage |
| Do you have an operations manager or developer who can own the rebuild? | Lean rip-and-replace | Lean ration-and-monitor |
| Is the usage clocked cost above £400 a month? | Act now (rebuild or ration) | Wait and re-test in six months |

That's the entire framework. Four questions and your response is largely determined.

## A worked example: the £1,200 to £180 move

The 30-person services firm that ran the rip-and-replace path is worth walking through in detail because it's the response that scares most SMBs and turns out to be the cheapest.

The starting point: HubSpot Marketing Hub Pro (£760 a month for the firm), plus HubSpot AI add-on usage (averaging £1,200 a month over Q1 2026). The AI was doing two main jobs. First, drafting outbound email sequences from a content brief. Second, summarising inbound leads from website forms and chat into structured CRM entries. The cost spike happened in February when the marketing team started using AI drafts on every email rather than only on net-new sequences.

We mapped the two workflows. Email drafting required a structured prompt and access to historical email tone. Lead summarisation required form data and a structured output schema. Both were within the standard capability range of Claude or GPT-4 class models without any vendor-specific tooling.

The rebuild took 14 hours of operations manager time over two weeks. We set up a Zapier workflow that takes a brief from the team's shared Notion, runs it through Anthropic's API, and posts the draft into HubSpot as an unsent email. We set up a separate workflow for lead summarisation that triggers on form submission, runs the form data through the API, and writes the summary back to the CRM. Total monthly API cost at the firm's volume came in at about £150, plus £30 for the Zapier plan, taking the total to £180. That's down from £1,200 a month at the same workload.

The kept HubSpot for everything else. Marketing Hub Pro is still doing the work it's good at: form tracking and sequences and reporting and list management. The firm just stopped paying HubSpot to run the AI inference layer when an external API was 85% cheaper.

That £12,240 annual saving is now redirected into the next AI workflow.

## What this means for your week

If you're paying for SaaS AI add-ons or watching them creep up on your monthly bill, do two things this week. First, get the actual usage report. Most SMBs don't know what their AI consumption looks like until they pull the data. Second, run the four-question framework above. Most SMBs will land in either ration-and-monitor or wait-and-re-test. A meaningful minority will discover they're well above the rip-and-replace threshold.

The "AI replaces SaaS" prediction was wrong about timing and wrong about mechanism. But the SaaS bill is still changing shape. So if you ignore the usage clock for the next two quarters, you may find your AI line item has doubled while your seat count stayed flat.

If you want a structured way to audit which SaaS AI features are worth keeping and which to replace, the AI Roadmap audit is the fastest path. It's at https://richardbatt.co.uk/roadmap. The Vault also has the SaaS AI audit template I use with clients, at https://richardbatt.co.uk/vault.

## Frequently asked questions

**Are HubSpot, Adobe and Salesforce really charging extra for AI now?**

Yes. The Information reported in late April 2026 (also covered by MarketScreener on April 27) that all three vendors are now charging customers based on AI usage. The pricing structures vary. Some are per-credit, some are tiered by feature, some are blended into seat plans with usage caps. The shared pattern is that AI cost is no longer included in the base seat price for most plans.

**Should I cancel my HubSpot or Salesforce AI features?**

Probably not by default. Run the four-question framework. The AI features are doing real work for many SMBs. The question is whether the usage clocked cost is justified for your usage. If your usage is consistent and over £400 a month, rip-and-replace usually pays back inside three months. Below that, ration-and-monitor or wait-and-re-test is often the right call.

**Can I really run the same workflows on raw OpenAI or Anthropic APIs?**

For most SMB workflows, yes. The gap between SaaS-bundled AI features and raw API access has narrowed significantly. The harder part isn't the model. It's the integration with your CRM, your email system, and your team's daily workflow. Tools like Zapier, Make and n8n close most of that gap without writing custom code. The example in this piece (£1,200 to £180) used Zapier and the Anthropic API and took 14 hours to build.

---

## More about Richard Batt

Richard Batt is an AI implementation specialist who helps businesses deploy working AI automation in days, not months. 120+ projects across 15+ industries.

### Key pages

- [Home](https://richardbatt.com/)
- [About Richard](https://richardbatt.com/about)
- [Blog](https://richardbatt.com/blog)
- [Contact](https://richardbatt.com/contact)
- [Subscribe](https://richardbatt.com/subscribe)

### Contact

- Email: richard@richardbatt.com
- Location: Middlesbrough, UK (working globally)
- Website: https://richardbatt.com