If you've started exploring automation for your business, you've probably come across two terms that seem to do similar things: Make (formerly Integromat) and AI agents. Both promise to save you time. Both involve connecting tools and automating tasks. But using the wrong one for the wrong job is like hiring a conveyor belt to do a chef's work — technically both move things along, but only one can actually cook. Understanding the difference between these two approaches will help you build automations that actually stick, save real money, and don't fall apart the moment something unexpected happens.
What Make Does (And Does Really Well)
Make is a workflow automation platform. You build visual "scenarios" — flowcharts that connect your apps and tell them what to do when something specific happens. When a new order comes in on Shopify, send a Slack message and update a Google Sheet. When a contact fills out a form, add them to your CRM and trigger a welcome email. It's rule-based, deterministic, and deeply reliable.
That reliability is its superpower. Make follows the exact same path every single time, which means you can trust it with high-volume, repetitive tasks where consistency matters. It integrates with over 1,500 apps, and most scenarios can be built by a non-developer in a few hours. Pricing starts at around £9/month for small usage, scaling with the number of "operations" you run.
The limitation is equally clear: Make cannot think. It can only do what you explicitly tell it to do. If a new edge case appears — a form submission with an unusual field, a customer reply that doesn't match the expected format — Make will either fail silently or throw an error. Every new variation requires you to go back in and build another branch.
A good rule of thumb: if you can write the exact steps on a sticky note without any "it depends," Make can probably handle it.
What AI Agents Do Differently
An AI agent is fundamentally different. Rather than following a fixed flowchart, an agent reasons through a task. You give it a goal — "qualify this inbound lead and draft a personalised follow-up email" — and it figures out the steps, pulls in context from your CRM, checks the lead's LinkedIn profile, evaluates their fit against your criteria, and writes a tailored message. No rigid flowchart required.
The key distinction is that AI agents handle ambiguity. They can read unstructured information (a messy customer email, a PDF contract, a voice note transcript), make judgements, and take actions that vary based on what they find. They're also capable of multi-step reasoning — breaking down a complex goal into sub-tasks and working through them sequentially.
This flexibility comes with trade-offs. Agents are more expensive to run than simple Make scenarios — depending on the model and usage, you might pay £0.05–£0.50 per complex task versus fractions of a penny for a Make operation. They also require more careful setup: you need to define the agent's goal clearly, give it the right tools to work with, and test it across varied inputs. And because they're probabilistic (meaning they don't always produce identical outputs), they're better suited to tasks where good judgement matters more than pixel-perfect consistency.
A Real Example: How a Consultancy Uses Both Together
A mid-sized management consultancy recently restructured their client onboarding process using both tools in tandem — and it's a useful illustration of how they complement each other.
Previously, onboarding a new client involved a consultant manually pulling together a welcome pack, creating project folders across three different tools, scheduling a kickoff call, and sending a series of emails over two weeks. The whole process took around four hours per new client.
They built a Make scenario to handle all the deterministic work: the moment a contract was signed in DocuSign, Make automatically created a project in Asana, set up a shared folder in Google Drive with the right template structure, added the client to their CRM, and triggered a calendar invite to the account manager. That alone saved roughly two hours per onboarding and eliminated three common errors (wrong folder names, missing CRM entries, forgotten calendar invites).
Then they layered in an AI agent to handle the variable parts. The agent read the signed contract to extract the client's specific deliverables, budget band, and industry, then drafted a personalised welcome email and a tailored two-week onboarding plan — something that genuinely required judgement and varied every single time. The account manager reviewed and sent it, but the drafting work — previously around 45 minutes per client — dropped to under five minutes.
Combined, they reduced onboarding admin from four hours to under 30 minutes per new client. At eight new clients per month, that's roughly 28 hours saved monthly — equivalent to nearly £2,000 in consultant time at their billing rates.
When to Use Each: A Simple Decision Framework
Rather than thinking of Make and AI agents as competitors, think of them as tools that sit at different points on a spectrum from "structured" to "unstructured."
Use Make when:
- The task always follows the same steps
- You're moving data between apps without needing to interpret it
- Volume is high and cost-per-operation matters
- You need 100% consistent, auditable outputs (financial data, compliance triggers)
- You can map the entire workflow on a whiteboard in under ten minutes
Use an AI agent when:
- The input varies and needs to be understood, not just routed
- The task requires judgement calls (prioritising, categorising, drafting, evaluating)
- You're dealing with unstructured content — emails, documents, transcripts
- The goal is clear but the steps to reach it aren't always the same
- Human review time is the bottleneck, not the volume of tasks
Use both together when:
- A workflow has predictable triggers and data routing (Make's territory) but also requires intelligent action on the content (the agent's job)
- You want to keep costs low on the structural work while reserving AI spend for the parts that genuinely need it
The consultancy example above is the ideal pattern: Make handles the plumbing, the agent handles the thinking.
Conclusion
Make and AI agents aren't rivals — they're different instruments in the same toolkit. Make is your reliable, low-cost workhorse for structured, high-volume tasks. AI agents are your flexible, reasoning-capable layer for everything that requires understanding context and making judgements. The businesses getting the most out of automation right now aren't choosing one over the other — they're learning where each belongs in their workflows and connecting them deliberately. Start by auditing one manual process this week: ask yourself whether every step is predictable. If it is, Make is your answer. If parts of it require someone to actually think, that's where an agent earns its keep.