Almost every business has one. A supplier portal, a partner dashboard, a carrier system, or a legacy internal tool with no API, where somebody logs in each morning, runs the same report, downloads a spreadsheet, and pastes it somewhere else. It is nobody's job and it takes twenty minutes a day, which is two working weeks a year for a task that requires no judgement at all.
Sessions are the recurring problem. Portals expire logins on unpredictable schedules, some rotate tokens per request, and a growing number sit behind multi-factor authentication that no automation should bypass. The workable pattern is a session manager that detects expiry and re-authenticates cleanly, combined with an escalation path for genuine MFA rather than an attempt to defeat it. Reports that generate asynchronously add a second problem, since the automation has to wait for readiness without a completion signal.
These systems are usually not defended, they are just awkward: a session that expires oddly, a report that takes ninety seconds to generate, a download that arrives as an email attachment rather than a file. The engineering is unglamorous and the return is unusually clear, because you can measure the hours saved against a fixed build cost and the arithmetic is normally obvious within a quarter.
The arithmetic is simple and worth doing before anything is built. Twenty minutes a day is roughly eighty hours a year. At a loaded cost of even a modest hourly rate, the annual cost of the manual version is usually several thousand dollars, against a one-off build in the low thousands. Most of these projects pay back inside a year, and the ones that do not are usually tasks that happen weekly rather than daily.
The cost that does not appear in that calculation is error and delay. A person who is on holiday does not run the report, so the downstream data is stale and nobody notices until a decision is made on it. Automation removes a dependency on someone remembering.
If the arithmetic does not work, I will say so. A task somebody does once a month for ten minutes is usually not worth automating, and I would rather tell you that by email.
Where the provider has an API, that is the build, and it is both cheaper and more durable than anything driving an interface. Checking for one is the first thing I do, and it is surprising how often a portal has an undocumented but supported API that nobody mentioned.
Where there is not, the automation drives the interface the way a person would, with a session manager handling authentication and expiry, explicit waits for asynchronously generated reports, and validation on the retrieved data before anything downstream consumes it. That validation matters: a portal that returns an empty report after a layout change should raise an alarm rather than silently overwrite good data with nothing.
Output goes wherever the person was pasting it, which is usually a spreadsheet, a database, or an accounting system. The point is to remove the manual step, not to introduce a new tool nobody asked for.
Workflow automation starts from £1,200 for one permitted supplier portal and one agreed report export. Authentication, scheduling, combined outputs and monitoring are assessed before quoting; a multi-portal system is a separate engagement.
Two things move the number. Whether an API exists, which can halve the work. And whether the portal uses multi-factor authentication, which sometimes means the schedule needs a human in the loop for re-authentication, or occasionally that full automation is not appropriate at all.
Portals change without notice and nobody tells the customer, so maintenance from £400 a month is sensible for anything a process depends on. The alternative is discovering it broke when someone asks why the numbers stopped updating.
Requires your own authorised account or API access. I do not build around logins you do not hold.
This work runs on accounts you already hold and are entitled to use, automating what your staff already do by hand. That is a materially different position from collecting data from a third party you have no relationship with. Two things still need checking, and I check them: whether your agreement with the provider permits automated access, since some explicitly restrict it, and whether automated retrieval could breach usage limits. Where a provider offers an API, using it is always better than driving the interface.
MathersLabs is not affiliated with or endorsed by Supplier and partner portals. Work is scoped to publicly accessible data and to compliance with applicable terms and law, and feasibility is assessed per engagement before any fixed price is agreed.
Brickwood
Five venues, one custom booking system, shipped in a week
Replaced a generic TypeForm with a purpose-built Next.js app for a five-venue South London café chain. Handles venue-specific menus, booking validation, full cost calculation, and routes on-brand confirmation emails to the right manager automatically.
AussieBankStatements
Bank statement converter and categorizer that never stores what it reads
A custom full-stack app on GCP that parses Australian bank statement PDFs into clean CSV. No AI, a dedicated parser per bank, and nothing written to disk. Built for people who want the extraction without handing their spending history to a chatbot that keeps it.
Tell me what data you need and where it lives. I will tell you whether it is feasible, what it would take, and whether it is worth building before you commit to anything. Working hours overlap with US Eastern and Pacific.