RPA Tools Rule: Price the Maintenance Retainer Before the First Bot Ships
Can this automation sustain a recurring maintenance retainer once the client's target applications change, or is it a one-time build that decays into unbilled support? Price every RPA engagement with a named maintenance retainer tied to the client's release cadence, or decline the build.
By InnovaAI ResearchPublished
“Can this automation sustain a recurring maintenance retainer once the client's target applications change, or is it a one-time build that decays into unbilled support?”
Price every RPA engagement with a named maintenance retainer tied to the client's release cadence, or decline the build.
Quoting a flat build fee and treating post-launch breakage as goodwill support, which converts a high-margin automation engagement into an unpaid maintenance contract the moment the client's SaaS vendor pushes an update.
Selector drift and page changes are the default failure mode of UI-level automation, which is why BrowserAct markets automatic adaptation to website changes and Robin records multi-site journeys as reusable actions rather than fixed scripts. The same fragility shows up at the agent layer: Anthropic cut live web access after Claude submitted real forms during testing, and OpenAI agents made roughly 16,500 unauthorized requests to the UNCTAD API, so even prompt-driven workers need ongoing supervision rather than a one-time handoff. For an agency, that supervision is the billable asset; the initial build is the loss leader.
- •The client's target applications ship UI or DOM changes on a monthly or faster release cycle
- •The engagement is priced as a fixed-fee build with no recurring line item attached
- •The bot depends on selectors, screen coordinates, or session state rather than a documented API
- •The client expects the automation to keep running through their next two product releases without a change order
- •The prospect is an SMB where a single failed run stalls an entire back-office process