Classic RPA bots to agentic automation
Cohash moves classic RPA estates to agentic automation in UiPath, or in Power Automate with Copilot Studio. We sort bots by stability and value and replace screen scraping with APIs wherever we can. Agent steps go in only where the work needs reading or judgment. Anything with money or access attached still needs a person to approve it. We also get UiPath versions back into support and move bot and agent credentials into a vault.
- 01Bot portfolio triage
- 02Platform currency
- 03APIs before agents
- 04Agent steps with guardrails
- 05Orchestration and evaluation
- 06Credentials and operations
The story
Told without the client's name or anything that could identify them.
A shared-services operations group ran a classic UiPath RPA estate whose bots broke whenever a target application changed a screen, while exception queues filled with emails and documents the bots couldn't read. Orchestrator 2023.10 was due to leave extended support on 23 October 2026, which forced a decision. We scored every bot first and decided which to keep, retire, replace with an API or extend with an agent step. Orchestrator and Studio moved to a supported LTS, screen scraping gave way to APIs where the targets offered them, and agents now do the reading work, with a person approving anything that touches money or access.
why it was urgentUiPath Orchestrator 2023.10 leaves extended support on 23 October 2026, and 2024.10 leaves mainstream support on 21 October 2026. Studio STS no longer edits Windows - Legacy projects. UiPath Maestro has been generally available since April 2025, and Copilot Studio agent flows can now take over Power Automate cloud flows.
- openBots break every time a target application changes a screen.
- recurringException queues are full of documents and emails the bots can't read.
- blockedOrchestrator or Studio is on a release that's out of support.
- openBot accounts share passwords stored in config files or the bot's own assets.
The estate, before and after
Before: screen-scraping bots and a full exception queue
Unattended bots drove ERP and CRM screens with shared passwords from config files. Anything they couldn't read landed with operations staff, and Orchestrator was heading out of support.
Read the diagram as text
- Email and documents → Unattended bots: pick up work
- UiPath Orchestrator 2023.10 → Unattended bots: schedule
- Shared bot accounts → Unattended bots: shared passwords
- Unattended bots → ERP: screen scraping
- Unattended bots → CRM: screen scraping
- Unattended bots → Operations staff: unreadable items
After: orchestrated robots, agent steps and human approval
Cases arrive in UiPath Maestro, which hands rule-based steps to robots and reading or judgment to agents. Anything uncertain lands with a person. Each bot and agent pulls its own scoped credentials from the vault.
Read the diagram as text
- Email and documents → UiPath Maestro: new case
- UiPath Maestro → Agent step: unstructured work
- UiPath Maestro → Robots on supported LTS: rule-based steps
- Copilot Studio agent flows → Agent step: AI action
- Agent step → Human approver: below threshold
- Credential vault → Robots on supported LTS: scoped credentials
- Credential vault → Agent step: scoped credentials
- Robots on supported LTS → ERP: API calls
- Robots on supported LTS → CRM: API calls
How we run it, phase by phase.
6 phases take you from Classic RPA bots to Agentic automation in UiPath or Power Automate. Each one ends with something you can check before the next starts. As you scroll, the blocks show the estate moving across.
- phase 01
Bot portfolio triage
Score every bot on run volume, failure rate, exception load and how stable its process is. Then make the call: keep it, retire it, replace it with an API integration, or add an agent step.
- phase 02
Platform currency
Upgrade Orchestrator and Studio to a supported LTS, convert Windows - Legacy projects to Windows projects, and retest the bots that stay.
- phase 03
APIs before agents
Swap UI automation for APIs or connectors where the target system offers them. That removes the most common break point before any AI is added.
- phase 04
Agent steps with guardrails
Bring agents in for the unstructured work, like reading documents, classifying emails or drafting replies. Spell out what the agent may do, set confidence thresholds, and send anything it isn't sure about to a person.
- phase 05
Orchestration and evaluation
Long processes run in UiPath Maestro or Copilot Studio agent flows. Keep test sets with known answers and check the agent's accuracy on each change before it's released.
- phase 06
Credentials and operations
Move robot and agent credentials to a vault, give each its own identity with narrow access, and track runs, costs and failures in one place.
What the client got
What changed for the client once the work was done. Described, not scored: we don't publish client figures.
- result 01
Each bot has a recorded decision
The portfolio scorecard has one row per bot, with the decision (keep, retire, API or agent) and the run volume, failure and exception numbers behind it.
- result 02
Fewer screens left to break
Where the target system offered an API or connector, the steps that used to scrape screens now call it. That took out the most common reason bots broke.
- result 03
Orchestrator and Studio are back in support
UiPath is back on a supported LTS. We converted the Windows - Legacy projects to Windows projects and retested the bots that stayed.
- result 04
Agents read, people approve
Agent steps classify emails and read documents within set confidence thresholds. Uncertain cases go to a person, and so do payments and access changes. Test sets with known answers rerun on every change.
- result 05
Every bot and agent has its own identity
We pulled the credentials out of config files and into a vault. The shared bot accounts were removed, and every identity has a named owner and narrow access.
Risks we plan for
-
Agents where rules would do
An agent adds variability. Steps that must give the same answer every time, such as posting a payment, stay deterministic.
-
Shared and over-privileged bot accounts
A single bot login shared across many processes defeats both audit and least privilege. Give every bot and agent an identity of its own, with an owner.
-
Capacity limits stop runs
If an environment runs out of Copilot Studio capacity, new agent flow runs just stop. Watch capacity and billing before critical work moves over.
-
Unsupported versions
Past extended support, nobody fixes Orchestrator and nobody answers when it breaks.
Where AI helps, and where it needs guardrails
UiPath Maestro orchestrates agents, robots and people in one process, and Copilot Studio agent flows put AI actions and human approvals into otherwise deterministic flows. The risky part is agents that act in ERP or CRM systems, directly or through MCP servers. They need scoped identities and logged actions, and a person signs off payments, access changes and anything that reaches a customer.
forward deployed engineers →What you're left with
We write these for your setup, and they're working documents. Your team should be able to keep things running from them long after we've gone.
- report01
Bot portfolio scorecard with a decision per bot
yours to keep - plan02
Upgrade plan to a supported UiPath LTS
yours to keep - design03
API replacement designs for UI-driven steps
yours to keep - design04
Agent step designs with thresholds and human approval points
yours to keep - tests05
Evaluation test sets and release checks
yours to keep - design06
Credential and identity model for bots and agents
yours to keep
Platforms and services on this move
Certified engineers, wherever you need them.
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.
Onsite
Engineers at your sites in the US
Nearshore
Overlapping hours with US teams
Offshore
Overnight progress on your backlog
All time zones
Coverage handed over across time zones
01Should we replace our bots with agents?
Mostly no. Stable bots that do rule-based work should keep running or become API integrations. Agents earn their place where people currently read, judge or write, such as invoices with odd layouts or customer emails.
02UiPath or Power Automate?
That depends on what you run now, your Microsoft licensing and how much attended or desktop automation you have. Teams with a big UiPath estate usually extend it. Microsoft-heavy shops with simpler flows often consolidate on Power Automate and Copilot Studio.
03How do you measure whether an agent step is good enough?
We build a test set from real past cases that have known correct answers. The agent is measured against it before release and after each change. Drop below the agreed accuracy and the step routes to a person.
Often on the same roadmap
AD FS federation to Entra ID authentication
Converting federated domains from AD FS to Entra ID cloud authentication, moving relying parties app by app and retiring the AD FS farm.
read the case studyAI pilot to production with forward deployed engineers
Taking an AI pilot into production with forward deployed engineers who work with the business team, build evals, connect real systems and hand over cleanly.
read the case studyAuth0 Rules and Hooks to Actions before end of life
Rewriting Auth0 Rules and Hooks as Actions before their 18 November 2026 end of life, with token claims compared before and after each change.
read the case study8fdd096a - 01UiPath docs: Product lifecycledocs.uipath.com ↗
- 02UiPath docs: About the Windows - Legacy compatibilitydocs.uipath.com ↗
- 03UiPath docs: Maestro release notes, April 2025docs.uipath.com ↗
- 04Microsoft Learn: Copilot Studio agent flows overviewlearn.microsoft.com ↗
Planning this move?
A 30-minute working session with a senior engineer. No slides.
- The platform and the version you're on
- Where you need to get to, and by when
- What's custom: rules, plugins, integrations
- Any audit finding or deadline driving it