Can one AI subscription replace 20 SaaS tools? The honest maths

"Replace 20 tools with one" is a great headline and a poor plan. Some vendors now advertise replacing 27 or more apps; Macro's homepage says exactly that about its unified workspace. The pitch has a real core: small businesses accumulate subscriptions that each solve a sliver of a problem, and the hidden cost is the person who copies information between them. But consolidating the wrong tool can cost more than it saves.

This piece belongs to our series on the AI execution layer, which works on top of your existing systems rather than replacing all of them. Here is how we would audit a stack, line by line.

A typical small-business stack, and what each line is really paying for

The table below is a composite of the stacks we see in Australian businesses with two to ten people. The spend column is an illustrative monthly budget in AUD, not a quote for any named product; your numbers will differ. The column that matters is the last one.

Add the rows and this illustrative business spends about $1,935 a month. The interesting part is the split. Roughly $500 is infrastructure and compliance that should never move. The rest, including the largest line, is paying for someone to do or chase work.

What genuinely consolidates

An agent subscription earns its place by absorbing tools whose whole job was to make a person faster at admin. In our experience these are the candidates:

• Reminder and confirmation tools. An agent that can call and text (in Cortex, through Fulcrum Lyre) can send confirmations and handle replies, not just fire a template.
• Meeting-note subscriptions. If an agent turns the conversation into tasks, drafts and CRM notes, a separate transcription tool adds little.
• Glue automations that keep breaking. Zaps that parse free-text emails or forms are the first to fail and the first to move. Clean, deterministic ones can stay; see Zapier vs AI agents.
• A basic CRM nobody updates. An agent-native CRM that agents write to as they work can replace a CRM that only existed on paper.
• Some of the VA hours. Not all. Agents handle repeatable admin; a person is still better at relationships and judgement calls.
• Simple website hosting and edits, if your plan includes hosting (Cortex's Small and Growing plans list website hosting).

What should not consolidate, even if a vendor says it can

Be stubborn about these. The cost of a mistake is far higher than the subscription:

• Accounting and payroll. Your accountant, the ATO and your bank feeds expect Xero or MYOB. Agents should work alongside them, preparing invoices and reconciliation for review, not replace the ledger.
• Trust accounting and regulated industry software. Property trust accounts, clinical records and NDIS claims have rules about where and how records are kept. Keep the certified system; let agents do the prep.
• Your email and identity provider. Google Workspace or Microsoft 365 is the backbone of access control. Agents should connect to it, not route around it.
• Tools your team genuinely loves. If the design team lives in a specific app, ripping it out to save $30 a month is a false economy.

Worked example: a five-person trades and service business

Using the illustrative stack above, here is what a realistic consolidation looks like. We assume the business moves to Cortex's Growing plan, listed at A$358 a month plus GST on the pricing page, keeps its infrastructure and industry software, and trims rather than cuts VA hours. Every figure apart from the Cortex plan price is illustrative.

Two honest observations. First, the licence saving is real but modest; for a business with a lean stack it might be close to zero. Second, the number that decides whether this is worth doing is not in the table: it is the owner's hours. If the owner was spending eight hours a week chasing, copying and updating, and gets five of them back, that is worth more than the subscription maths. Measure it properly before you decide, using our method for measuring hours saved by AI agents.

The order to switch tools off

Never cancel first. Run the agent alongside the old tool, compare results, then cancel. The order that minimises risk:

1. Glue automations that break most often. They are already costing you time.
2. Reminders and confirmations. Easy to compare side by side for two weeks.
3. Meeting notes and follow-up drafting. Low risk; you review every output anyway.
4. Lead research subscriptions. Compare list quality, not list size.
5. The CRM, only after agents have been writing to the new one for a month and the data is better than the old one.
6. VA hours, last, and gradually. Re-scope the role towards work people do best.

Hidden costs of consolidating

• Export and migration time. Getting clean data out of an old CRM is usually the longest step. Budget for it.
• Review time. Agents draft; you approve. Ten minutes a day is normal at the start and is part of the real cost.
• Concentration risk. One subscription doing twenty jobs is one vendor you depend on. Have an exit plan before you need it; our AI vendor lock-in checklist covers data export and portability.
• Change fatigue. Switching four tools in a month exhausts a small team. Two at a time is plenty.

When you should leave your stack exactly as it is

If your tools are well integrated, your team uses them properly and the CRM is genuinely up to date, an AI subscription should be justified by new capacity (calls you weren't making, research you weren't doing), not by cancellations. And if your problem is that nobody follows the process, consolidating software will not fix it. Fix the process, then automate it.

If you want to model your own stack, list every subscription with what it is really paying for, then compare against the plans on our pricing page. Bring that list to a demo and we will tell you which lines we think should stay.