What we built this week

Every week, five short posts in plain English about what changed in Automize. Scroll down for earlier weeks.

17 to 23 October 2026

Bot health checks, smarter selectors, marketplace, and tighter Production controls

This week we shipped machine health and readiness checks so a bot warns you before a run fails, a Smart selector that finds a screen element from plain English when the saved one breaks, a rebuilt marketplace with working search and import from UiPath and n8n, tighter Production environment controls so only promoted processes run there, and a fix that lets you run a promoted process from the launchpad without errors.

Your bot now warns you before a run fails, not after

When a bot stops working overnight, the first sign is usually a failed run the next morning. By then the work is already late.

We built a health and readiness check that runs before each process starts. The bot now reports its clock drift, its own memory use, leftover Excel and Word copies, sleep and lock settings, and whether it has the software a process needs. If something looks wrong, a warning appears on the bot page before anyone presses Run.

For unattended bots, the page now flags three problems clearly: Windows not set to sign in by itself, the bot not set to start with Windows, and the bot not set to restart if it stops. Any one of those means a trigger will fire and nothing will happen. The page also shows a "How it's set up" card so the settings that matter are no longer buried under Take control.

The Play window lists machines that are ready for a process first, and a machine that is missing something the process needs says so plainly instead of letting the run start and fail.

#Automize #DigitalWorkers #Automation #BotHealth

A broken screen selector now heals itself from a plain-English description

A bot step breaks when the screen changes and the saved selector no longer finds its element. Fixing it means opening the process, finding the step, re-recording the selector, and promoting the change. That takes time nobody has when a run is waiting.

We added a Smart selector: one sentence in plain English that describes the element, for example "the text box to the right of the label Name". The bot tries every saved selector first. Only if all of them fail does it use the sentence, first without AI and then with it, to find the right element on screen.

When Smart finds a match, it sends the result back to the web for a person to review and accept. It never promotes a fix to Production on its own. The step form shows the Smart sentence as the last row in the element's fallback list, clearly marked, so it is obvious what it is and easy to clear if the element itself has changed.

This feature is on by default and counts against a company's AI allowance. It can be switched off, or pointed at a company's own AI connection.

#Automize #Automation #RPA #SmartSelector

Import a UiPath or n8n workflow and find ready-made processes in the marketplace

Starting a new automation from scratch takes time. Most teams have already built something similar in another tool, or someone in the community has.

The marketplace now lets you import a workflow you already have. A UiPath project file or zip becomes a bot process. An n8n export becomes a cloud process. Steps that map directly are converted. Steps with no match come in as placeholders with a note on what to finish. We measured this on nearly 600 real public workflows and the number that import with no placeholders at all roughly quadrupled.

The marketplace itself now loads properly. Featured items, category filters, trending lists, version tabs and related items were all blank before because the list responses were dropping their rows. That is fixed. Category chips only appear for categories that have items, and the count no longer includes items that have been switched off.

The import note, source link and licence details are kept for staff but are not shown to customers browsing the marketplace.

#Automize #Automation #Marketplace #RPA

Production now only runs what was promoted, and bots stay in their own environment

Before this week, a bot set up for Production could pick up a Development process, and editing a promoted process reached Production straight away with no approval. That made it hard to know exactly what was running in Production at any moment.

We tightened this up. A bot now only takes work from the environment it is set up for. Production only ever runs a frozen, promoted version of a process, never the draft you are editing. Promoting is the one path to Production, and a schedule, webhook or queue cannot run an unpromoted process there.

The bot's Processes tab now shows each process's environment. A process assigned to a Production machine that has never been promoted there gets a red badge and a plain note on how to fix it. Process cards show the environment in red when it is Production, so the status is visible at a glance.

Processes already running in Production before this change keep running. The process form no longer offers Production as a starting environment for new processes.

#Automize #Automation #Production #ProcessManagement

Savings reports now count by work item, and client emails show time and money saved

A single failed run could make a whole batch of work look like a failure, even when most items completed successfully. That made it hard to show clients or managers what automation was actually delivering.

Process performance and bot reliability now count by work item when a process records them. An item that was retried and then completed counts as a success. One failed item no longer drags down the whole run's result. Runs remain the measure for processes that do not record individual items.

Client emails can now include a "Time and money saved" section. When building an email you choose the period, whether to add running totals for the week, month or year, and which figures to show: work done, time saved, cost of a person, and people freed up. The calculation is the same one the dashboard uses. Processes with no manual time set are named so they can be updated. Test runs are left out of savings figures everywhere, so numbers from while a process was being built do not inflate the results.

#Automize #Automation #ProcessPerformance #Reporting

Get started

See it working on your own process

The 7-Day Proof-of-Value builds one working automation on your real data, at a fixed price, with an ROI report at the end.