Agents, workflows or custom software: picking the right automation

“We should use AI for this” is a good instinct and a bad brief. The same frustration, whether it’s leads going cold, reports eating up every Friday or a booking process held together by spreadsheets, can call for very different fixes.

Most of what we build takes one of three shapes: a ready-made agent, an AI workflow engagement or a piece of custom software. Here’s how to tell which one you need.

Start with the job, not the tool

Before choosing anything, write the job down in one sentence: what comes in, what should happen and what “done” looks like. “Every new lead gets a call and a booked meeting” is a job. “Use AI in sales” is not.

Then ask two questions. Is this job common to lots of businesses, or specific to how yours works? And does it need a new place for people to log in and work, or should it run in the background across the tools you already use?

When a ready-made agent fits

Some jobs look much the same everywhere: calling and qualifying leads, finding and scoring prospects, writing cold email, keeping a CRM clean, and producing SEO content, ads and social posts. For these, a specialised agent is the fastest route. We run nine of them across sales and marketing, each one taking a single slice of manual work off your plate and running it around the clock.

A ready-made agent is usually the right call when:

• Your processes are already mapped and clear.
• You know which workflow to automate first.
• You want it live in weeks, not months, on the tools you already run.

Ready-made doesn’t mean generic. Each agent is still tailored to your data, connected to your tools and tested against real inputs before it goes live. But the shape of the job is known, so there’s far less to design.

See every agent, and what each one takes off your plate, on our products page.

When you need an AI workflow engagement

Other jobs are tangled up in how your business specifically operates: approvals that pass through several people, data spread across systems, edge cases only your team understands. Or you know AI could help but can’t yet say where to start. The signs:

• You’ve tried tools that didn’t stick.
• Your processes live in people’s heads, not docs.
• You want an ROI-ranked roadmap, not a sales pitch.

This is where a strategy-first engagement earns its keep. We document how the work actually happens today, rank every automation candidate by impact, effort and ROI, pick the stack and write the SOPs, then build, deploy and run the agents that do the work. You can also take the roadmap to your own team.

Picking the platform underneath

Half of this conversation is tooling, and the wrong platform choice is the most expensive mistake we fix. As a rough guide:

• Fast no-code automations: Zapier or Make, for broad integrations and quick deployment.
• Complex, self-hosted AI workflows: n8n, for full control and far better pricing at scale.
• Multi-agent AI pipelines: LangChain or LangGraph, purpose-built for agent orchestration.
• Team communication and alerts: Slack or Discord, for native, low-friction integration.

Our AI agent automation service covers this end to end, from the audit to agents running in production.

When custom software is the answer

Sometimes the problem isn’t a missing automation. It’s a missing system. Watch for these signs:

• Sticky notes and spreadsheets. You’re tracking things by hand because nothing was ever built for exactly how you work.
• Too many apps. Three tools and a spreadsheet, stitched together, doing the job of one.
• It broke when real customers showed up. It worked fine in the demo, then fell apart once real people started using it.

Here, putting an agent on top of the mess just automates the mess. What you need is a booking system, a client portal or a dashboard built around how you actually work, connected to the tools you already run and owned outright by you: the code, the data and the infrastructure. Most of these projects go from idea to launch in 8 to 12 weeks, depending on scope.

That’s our full-stack application development service.

They’re not mutually exclusive

The three often stack. A custom booking system can feed a voice agent that confirms appointments. A strategy audit can end with a ready-made agent as step one of the roadmap. What matters is the order: fix the system if it’s broken, design the workflow if it’s unclear, and reach for a ready-made agent when the job is well understood.

Still not sure which shape fits? That’s what the first call is for. Bring the job, and leave with a clear first step.