Of your two hundred bots, forty matter. Nobody can tell you which ones the next app release will break.
RPA programs grow one bot at a time. After a few years there are unattended robots sharing credentials and screen-scraping flows that fail with each vendor update, and now agents are being added on top. We work out what you have first. Then we build new automation on UiPath, Power Automate or ServiceNow.
Robots and desktop flowsUiPath · Power AutomateCredential vaultper-bot identities
04 · Systems
S/4ERP and financeSAP · legacy appsService platformServiceNow RPA Hub
UiPathMicrosoft Power PlatformServiceNowCyberArk
in short·20-second read
Intelligent automation uses robotic process automation, workflow orchestration and AI to run business processes with less manual work, usually on UiPath, Microsoft Power Automate or ServiceNow.
Cohash takes stock of existing bot estates and stabilizes them, replacing fragile screen automation with APIs where we can. We build new automations and agentic workflows such as UiPath Maestro processes. We also set up credential and governance controls and staff certified automation developers.
you probably need this if
Nobody can say how many bots are running or who owns each one.
Bots fail after each update to the applications they automate.
Several of your robots share one service account login, and its password sits in a config file.
From the first call to a setup your own team can run.
Here's what usually prompts a call about your Intelligent Automation, what we do about it and in which order, and where you end up.
01 · where it hurts02 · what we change03 · how it runs04 · where you end up
incident board · intelligent automation3 open
openNobody can say how many bots are running or who owns each one.
recurringBots fail after each update to the applications they automate.
blockedSeveral of your robots share one service account login, and its password sits in a config file.
exposure
Bot estate inventory and cleanupUiPath development and…Power Automate desktop flowsServiceNow automationCredentials and governanceCertified automation staffing
S/4
startrehearsehandover
Inventory and triage
Stabilize what matters
Build new work
Run and govern
we are here
ready to hand over
Most invoice steps no longer care what the screen looks like.
Each robot's actions are traceable to its own identity, credentials rotate without code…
Routing now comes from a step you can measure and improve.
Automation inventory with owner, value, failure rate and decision
chapter 01 · where it hurts
Why teams reach out
Nobody can say how many bots are running or who owns each one.
Bots fail after each update to the applications they automate.
Several of your robots share one service account login, and its password sits in a config file.
chapter 02 · what we change
How we solve it
Bot estate inventory and cleanup
UiPath development and orchestration
Power Automate desktop flows
ServiceNow automation
Credentials and governance
Certified automation staffing
chapter 03 · how it runs
Our approach
Inventory and triage
Stabilize what matters
Build new work
Run and govern
chapter 04 · where you end up
What changes
Most invoice steps no longer care what the screen looks like.
Each robot's actions are traceable to its own identity, credentials rotate without code changes, and developers no longer see production passwords.
Routing now comes from a step you can measure and improve.
Automation inventory with owner, value, failure rate and decision
what we do
Where we spend the hours
6 areas of work, each one done by an engineer who has run it in production.
We list every robot, flow and schedule, with its owner, what it's worth to the business, how often it fails and which applications it touches. Then each one gets a decision. Keep it, fix it, rebuild it on an API, or retire it.
Unattended and attended automations, queue-based processing, and agentic processes coordinated with UiPath Maestro, where AI agents, robots and people each handle the steps they're suited to.
Desktop flows for Windows and legacy applications, including terminal emulators, with cloud flows for orchestration and process mining to find which processes are worth automating first.
Run IT and employee service processes on the platform where the tickets already live. Use Integration Hub spokes, flows built in Workflow Studio, and ServiceNow RPA Hub for desktop steps.
Move robot accounts out of config files and into a vault. Give each bot one identity with least privilege. Run a center-of-excellence intake process, and apply change controls when a target application updates.
Automation developers for UiPath, Power Automate and ServiceNow, plus solution architects and support engineers. Contract, contract-to-hire or permanent.
capability 01 / 06in production
S/4
Bot estate inventory and cleanup
We list every robot, flow and schedule, with its owner, what it's worth to the business, how often it fails and which applications it touches. Then each one gets a decision. Keep it, fix it, rebuild it on an API, or retire it.
Unattended and attended automations, queue-based processing, and agentic processes coordinated with UiPath Maestro, where AI agents, robots and people each handle the steps they're suited to.
Desktop flows for Windows and legacy applications, including terminal emulators, with cloud flows for orchestration and process mining to find which processes are worth automating first.
Run IT and employee service processes on the platform where the tickets already live. Use Integration Hub spokes, flows built in Workflow Studio, and ServiceNow RPA Hub for desktop steps.
Move robot accounts out of config files and into a vault. Give each bot one identity with least privilege. Run a center-of-excellence intake process, and apply change controls when a target application updates.
Automation developers for UiPath, Power Automate and ServiceNow, plus solution architects and support engineers. Contract, contract-to-hire or permanent.
We pieced these together from Intelligent Automation situations that come up a lot. They're patterns. You won't find a specific client in them.
case file 01
Invoice bots that fail every time the ERP screen changes
composite · not a client story
Beforethe situation
Unattended robots key invoices into the ERP through its screens. Every ERP update moves a field or adds a dialog box, and the bots stay broken until a developer re-records the selectors.
What we dothe work
Check which steps the ERP exposes through APIs and rebuild those first, keeping UI automation only where no API exists. Add a test run against the ERP test environment as part of every ERP release.
Afterwhere you end up
Most invoice steps no longer care what the screen looks like. The few that still do get tested before each ERP release, instead of failing after it.
case file 02
Shared robot credentials in a config file
composite · not a client story
Beforethe situation
A dozen unattended robots run under one service account with broad ERP and email permissions. The password sits in an orchestrator asset that several developers can read.
What we dothe work
Create one identity per robot, with only the permissions its process needs. Move credentials into the vault and turn on rotation. Update the orchestrator so it fetches them at runtime.
Afterwhere you end up
Each robot's actions are traceable to its own identity, credentials rotate without code changes, and developers no longer see production passwords.
case file 03
An agent bolted onto an unreliable email triage bot
composite · not a client story
Beforethe situation
A robot reads a shared mailbox and routes requests using keyword rules that misroute many emails. A team wants to add an AI agent to fix routing without changing anything else.
What we dothe work
Put the AI classification step in front of the routing, keep the robot for system updates, send low-confidence emails to a person, and track routing accuracy against a labeled sample.
Afterwhere you end up
Routing now comes from a step you can measure and improve. Robots do only the system work. When an email is uncertain, a person decides.
How robots, agents and people share an automated process
What to automate comes out of process mining and intake. The orchestrator gives AI agents the judgment steps, gives robots the system work and leaves approvals to people. Robot credentials come from a vault. Results go back into the business systems.
01 · Pick the work
Process miningevent logs · variants → Automation intake · candidates
OrchestratorUiPath Maestro · cloud flows → AI agent steps · judgment step→ Human approvals · exception→ Robots and desktop flows · system step
AI agent stepsclassify · extract · draft
Human approvalsexceptions · sign-off → Orchestrator · decision
03 · Execute
Robots and desktop flowsUiPath · Power Automate → ERP and finance · API or UI→ Service platform · update ticket
Credential vaultper-bot identities → Robots and desktop flows · credentials
04 · Systems
S/4
ERP and financeSAP · legacy apps
Service platformServiceNow RPA Hub
Read the diagram as text
Process mining → Automation intake: candidates
Automation intake → Orchestrator: approved process
Orchestrator → AI agent steps: judgment step
Orchestrator → Human approvals: exception
Orchestrator → Robots and desktop flows: system step
Credential vault → Robots and desktop flows: credentials
Robots and desktop flows → ERP and finance: API or UI
Robots and desktop flows → Service platform: update ticket
Human approvals → Orchestrator: decision
how it runs
A typical engagement
We don't start a step until the last one has produced something you can check.
step 01
Inventory and triage
List every automation. Measure how often each fails and what it's worth. Find the ones that depend on screen scraping an API could replace.
step 02
Stabilize what matters
Start with automations that carry the most value and fail most often. Fix them or rebuild them, and put their credentials in a vault.
step 03
Build new work
New automations and agentic processes chosen from process mining data and business intake, built with tests and monitoring.
step 04
Run and govern
We hand over monitoring, a support rota and an intake process, plus a release check that runs the bot tests whenever a target application changes.
you keep
What you're left with
We write these for your Intelligent Automation setup, and they're working documents. Your team should be able to keep things running from them long after we've gone.
inventory01
Automation inventory with owner, value, failure rate and decision
yours to keep
plan02
Rebuild plan for fragile automations, API-first where possible
yours to keep
design03
Credential vault integration and per-bot identity design
yours to keep
tests04
New automations with test cases and monitoring
yours to keep
evidence05
Center-of-excellence intake and change control process
Some clients want a project team, some want a couple of engineers on theirs. Either works. Onsite, nearshore or offshore, with handovers across time zones that keep things moving.
Put these to anyone bidding on your Intelligent Automation work. The ones who've done it answer differently from the ones who've read about it.
Q1
How will you decide which of our existing bots to retire?
Large bot estates usually have low-value, high-maintenance automations in them. If the partner only ever rebuilds, your support bill goes up.
Q2
Which UI steps would you replace with APIs, and how do you find them?
Most bot failures come from screen automation. Ask how the partner checks each target application for usable APIs.
Q3
How will robot credentials be stored, scoped and rotated?
Shared bot accounts with broad access turn up in audit findings all the time. Expect per-bot identities and a vault.
Q4
How do you test bots when a target application is updated?
With no release check, a vendor update breaks bots in production. Ask how the bot tests hook into your application change process.
Q5
Where would you use an AI agent, and how would you measure it?
Agents help with judgment steps but add uncertainty. The partner should name the step, the fallback and the accuracy measure.
01Should we use UiPath, Power Automate or ServiceNow?
It depends on what you already license, where the processes run and who'll maintain them. Power Automate fits Microsoft 365-heavy desktops. ServiceNow fits processes that start and end in tickets. UiPath fits large cross-application estates with heavy unattended workloads. Lots of organizations run more than one.
02Why do our bots keep breaking?
Most failures come from screen automation that depends on UI elements a vendor changes, expired or shared credentials, and unexpected data. Replacing UI steps with APIs where they exist removes the biggest source.
03Where do AI agents fit with RPA?
Give agents the judgment steps, such as reading an unstructured email or classifying a document. Leave deterministic system work to robots. UiPath Maestro and similar orchestration tools coordinate both, while people approve exceptions.
04How long does a bot estate cleanup take?
It depends on the number of automations, how well they're documented and how many target applications they touch. The inventory usually takes a few weeks, and the fix list sets the rest.
05Can you support our bots after go-live?
Yes. We provide support with monitoring, runbooks and a release check against application updates, with coverage across time zones if you need it.
06Can you provide UiPath developers on contract?
Yes. We place certified automation developers for UiPath, Power Automate and ServiceNow through our staffing practice. Contract, contract-to-hire or permanent.
related
Often paired with
Projects on Intelligent Automation usually end up touching one of these, so they're worth a look.
