Processes that ran
on human time
now run on their own

Four things make a process a strong automation candidate: it's high-volume (at least 50 repetitions per week), it follows consistent rules most of the time, it uses digital inputs and outputs, and it currently produces errors or causes delays. Invoice processing, employee onboarding, compliance reporting, and data migration between systems typically score well on all four. We run a free 90-minute process assessment call to score your specific process before any engagement begins.

RPA replicates what a human does on a screen - clicking, copying, pasting - without changing the underlying systems. It's fast to deploy and works well for stable, rule-based processes. Workflow automation orchestrates multi-step processes across systems via APIs, with conditional logic, human approval steps, and state management. Intelligent automation adds an AI layer - NLP, document understanding, classification - for processes that involve unstructured data. Most of our projects combine all three, starting with the simplest layer that solves the problem and adding intelligence only where rules alone can't handle the variation.

A focused single-process RPA implementation - one bot, one workflow, one set of systems - typically takes 4–8 weeks from discovery to production. A multi-process automation programme with orchestration, exception handling, and a monitoring dashboard is typically 10–18 weeks. The variable that affects timeline most significantly is process stability: if the process has a lot of undocumented edge cases that only surface during testing, scope expands. Our discovery phase is designed to find these before development begins.

This is the most common reason automation fails in practice, and it's a maintenance question as much as a build question. We address it by building bots with configurable selectors rather than brittle hardcoded element IDs, adding health checks that alert before a failure cascades, and including a 60-day hypercare period after go-live specifically to catch UI changes that happen in the weeks after deployment. Every automation we deliver includes a runbook for common failure scenarios and an SLA for response time when a bot goes down.

Almost never. RPA works specifically because it interacts with systems at the UI layer rather than requiring API access. We automate processes that span legacy ERP systems, web portals, email, spreadsheets, and modern SaaS tools without touching the underlying architecture. The exception is when a process involves a system that's genuinely too unstable to automate reliably - in which case we'll tell you before we start, not after we've built the bot.

A single focused RPA bot typically costs between $18,000 and $45,000 to design, build, test, and deploy, depending on complexity and the number of systems involved. A multi-bot programme with orchestration infrastructure is typically $80,000–$200,000. Based on our project history, most automation investments pay back within 5–9 months at the volume levels that make them worth automating. We model this in detail during the feasibility phase - you'll have a ROI projection with conservative and realistic scenarios before we agree to build.

We offer two models. The first is a clean handover with full documentation, bot health monitoring configured, and a runbook - your IT team or operations team takes ownership with 60 days of hypercare from us. The second is an ongoing managed service retainer where we monitor bot health, handle maintenance when upstream systems change, and add new processes to the automation programme on a monthly basis. About 60% of clients choose the retainer model in year two after seeing how much the bots are doing.