AI & automation
AI workflow automation means a process that used to need a person to read, sort, copy or chase now runs on its own, with a language model doing the one step a fixed rule cannot, such as reading a messy email or classifying an inbound WhatsApp message. We build these workflows with tools such as n8n, Make or Zapier, or in custom code when the workflow is core to your business. We start with one process, record how it runs today, and measure again after it has run for a while.
A common pattern: a team buys a tool, connects ten things in the first week, and finds a month later that nobody trusts the output. Nobody wrote down how long the manual process took, so nobody can say if it improved. The automation touches the CRM, the inbox and the spreadsheet, and when one record goes wrong it is hard to tell which step caused it.
We work the other way round. One workflow, with a named owner on your side. We time the manual version first. A fixed rule handles whatever a rule can handle, and a model is added only for the judgment calls. For the first weeks a person reviews what the workflow did before it is allowed to act alone.
Leads, support emails and WhatsApp messages read, classified and routed to the right person or queue, with the original message attached so nobody has to ask again.
Fields pulled from invoices, forms, PDFs and email threads into your CRM, sheet or database, with unclear items flagged for a person instead of guessed.
Records created and updated across your tools, so the data entry that used to follow every call or enquiry happens by itself.
Weekly digests, call summaries and status reports assembled from your own data and sent where your team already looks.
For teams that get enquiries from forms, email, ads and WhatsApp and lose time sorting them. A message arrives, a model classifies it (new enquiry, existing customer, spam, complaint), and the workflow creates the lead, assigns an owner and drafts a reply for a person to approve. We already build lead routing and WhatsApp Cloud API plumbing for BrokLeads, so the connection work is familiar ground.
How BrokLeads routes WhatsApp leadsInvoices, application forms, delivery notes and long email threads, turned into structured rows. Every extracted field carries a flag for how sure the workflow is, and anything below the line goes to a review queue. The review queue is part of the design, not an afterthought.
Every AI workflow automation project runs through our delivery process. These are the steps that are specific to this work. Teal steps are thinking, graphite steps are building, and orange ones keep it running.
Choose the workflow
We list the processes that repeat most, rank them by frequency and pain, and pick one. The rest wait.
Record the baseline
We time the manual version, count the manual touches and note how often it goes wrong, so there is something to compare against.
Design the steps
A diagram of every step, which are fixed rules, which need a model, where a person reviews, and what happens when a step fails.
Build and connect
The workflow built in the tool that fits (a platform such as n8n, Make or Zapier, or custom code), connected to your CRM, inbox or database.
Run with human review
For the first weeks a person approves what the workflow does. We read the exceptions together and tighten the rules.
Measure and decide
After about a month we compare against the baseline, report what changed, and decide whether to let it run alone, adjust it or stop.
| Scope | Range |
|---|---|
| Pick the workflow and record the baseline | 1–2 weeks |
| One workflow built, connected and in supervised use | 3–6 weeks |
| Several linked workflows with custom code and reporting | 2–4 months |
Planning ranges, not quotes. Your scope sets the real timeline, which we confirm in writing after discovery. See how engagements are priced.
When the job is a conversation with customers rather than a background process.
If you have many candidate workflows and want help choosing before anything is built.
The channel and routing plumbing we already run in production.
How we structure the code behind integrations when a no-code tool is not enough.
The aim is to remove the sorting, copying and chasing, not the judgment. In the workflows we design, a person still reviews anything uncertain or consequential. Whether you then use the freed time for growth or restructure is your decision, and we won't present it as a given.
What gets sent to a provider is a decision we make with you: we send only the fields the step needs, mask what we can, and review the provider's retention and training settings together. We design with data minimization in mind. Your own legal review decides whether a given provider is acceptable for your data, and we don't claim any compliance certification.
We won't quote an accuracy number before we've seen your real messages and documents, because it depends on how clean and varied they are. We test on a sample of your data first. Uncertain items go to a review queue instead of being acted on, every run is logged, and risky actions such as sending to a customer wait for approval until the record shows they are safe to automate.
A named person who knows the process, examples of real inputs (emails, forms, documents, with sensitive details handled as agreed), access to the tools involved, and a few hours of review during the first weeks. Without those, we'd be automating a guess.
We record the manual time, the number of manual touches and the error rate before building, then measure again after about a month. Running costs come from the automation platform, model usage and some monitoring time; we estimate them with you during scoping, and they depend on volume and the tools you choose.
You can, and for a simple workflow that may be the right call; we'll say so. We add value when the workflow touches several systems, needs reliable error handling and logging, handles private data, or has outgrown a chain of fragile zaps. We're developers, so when a no-code tool hits its limit we can write the code instead.
Chatbots and agents for your website and WhatsApp, with a narrow scope, a human handover and guardrails.
Learn moreCRMs, internal tools, portals and multi-tenant SaaS with roles, permissions and integrations.
Learn moreUse-case selection, data audit, build-or-buy and a written pilot plan, without hype.
Learn moreA few lines is enough. We'll ask the right questions on the call.
We reply within one working day
By email or WhatsApp, whichever you prefer.
A short call
To understand the goal, the people who'll use it and any deadline.
A written scope and estimate
Before any work starts, so you know what you're agreeing to.
Prefer to talk first?