By 2026, Robotic Process Automation (RPA) is entering a new phase. Traditional software robots remain highly relevant for structured, rule-based tasks, but enterprises are increasingly combining RPA with generative AI, AI agents, process orchestration and intelligent decision-making.

The result is a transition from "bots executing tasks" to "AI-driven systems managing processes."

Gartner's 2026 research describes RPA as still particularly effective for reliable UI-based task automation, while identifying computer use, agentic automation, and business orchestration as major forces changing the market.

1. From Traditional RPA to Agentic Automation

Traditional RPA follows predefined rules:

If X happens → execute Y.

This works extremely well for predictable processes such as:

• copying data between applications
• invoice processing
• account reconciliation
• employee onboarding
• report generation
• data validation
• legacy-system integration

The new generation of automation introduces AI agents capable of interpreting context, selecting actions and adapting to changing conditions.

Instead of:

Trigger → Bot → Fixed sequence → Result

the emerging model becomes:

Business goal → AI agent → Reason → Use tools → Execute → Validate → Escalate if necessary

Gartner identifies agentic automation as one of the major emerging areas in enterprise automation.

This does not mean that traditional RPA disappears. Instead, RPA becomes one execution capability within a broader AI automation architecture.

2. RPA + Generative AI

One of the most important developments is the integration of RPA with generative AI.

Traditional RPA struggles with unstructured information.

For example, a bot can easily read:

Customer ID = 123456

but has difficulty understanding:

"The customer appears to have been charged twice. Please investigate the previous invoices and determine whether a refund is required."

Generative AI can interpret the request, analyse documents and determine the next step.

RPA can then execute the transaction.

This creates a powerful combination:

Generative AI = understand and reason

RPA = execute

APIs = integrate

Orchestration = coordinate

Human = supervise and handle exceptions

Deloitte similarly describes the combination of AI agents and RPA as a way to extend automation from structured tasks into more dynamic processes.

3. AI Agents Become the New Automation Layer

The biggest conceptual change is the emergence of agentic automation.

An AI agent can potentially:

1. receive a business objective
2. understand the context
3. retrieve information
4. decide which systems to access
5. execute actions
6. evaluate the result
7. retry or choose another approach
8. escalate exceptions to humans

For example, imagine an SAP procurement process.

A traditional RPA bot might:

Read purchase order → enter data → click approval → update system

An AI-enabled process could:

Understand purchasing request → check supplier → analyse historical prices → verify budget → determine required approval → update SAP → communicate with supplier → monitor completion

This represents a significant change in the role of automation.

The automation is no longer simply executing instructions.

It is increasingly managing a business objective within defined guardrails.

4. Computer Use Will Reduce the Dependence on APIs

Another important trend is AI-driven computer use.

Historically, enterprise automation often required:

• APIs
• connectors
• database access
• custom integrations
• RPA screen automation

AI systems are increasingly capable of interacting with software interfaces in a more human-like way.

This is particularly relevant for organisations operating large landscapes containing:

• SAP
• legacy applications
• mainframes
• proprietary systems
• web applications
• desktop applications

Gartner specifically identifies computer use as one of the technologies influencing the evolution of the RPA market.

This could be particularly important for legacy transformation.

Instead of spending months developing an integration, an AI-enabled automation layer may be able to interact with an existing application while the organisation gradually modernises the underlying architecture.

5. Process Orchestration Becomes More Important

The future of automation is not simply about creating more bots.

It is about coordinating people + bots + APIs + AI agents + applications.

Consider a customer-service process:

Customer request

↓

AI interprets request

↓

Check CRM

↓

Check SAP

↓

RPA accesses legacy system

↓

AI evaluates result

↓

API executes transaction

↓

Human approves exception

↓

Customer receives response

This requires an orchestration layer.

The emerging architecture therefore looks increasingly like:

AI Agents

↓

Process Orchestration

↓

RPA + APIs + Applications + Data

↓

Human Workforce

Camunda's 2026 research highlights orchestration as a key requirement for moving from isolated AI experiments toward enterprise-scale agentic automation.

6. From Task Automation to End-to-End Process Automation

Traditional RPA often focuses on individual tasks.

The new generation focuses increasingly on the entire business process.

For example:

Traditional approach

Automate:

• invoice entry
• invoice validation
• approval notification

New approach

Redesign:

Invoice received → understand → validate → match → investigate exceptions → approve → post to SAP → pay → reconcile → report

This distinction is extremely important.

Adding AI to a poorly designed process does not necessarily create transformation.

The organisation must ask:

Should we automate this process as it exists, or should we redesign the process first?

BCG's 2026 research similarly argues that end-to-end process redesign is an important differentiator in achieving significant value from agentic automation.

7. Intelligent Document Processing Becomes Standard

Another major trend is the combination of RPA, OCR, computer vision and generative AI.

Traditional OCR extracts text.

AI-powered document processing can increasingly:

• understand document structure
• classify documents
• extract information
• interpret meaning
• identify inconsistencies
• compare multiple documents
• detect missing information
• recommend actions

For example:

Purchase order + invoice + delivery note

can be analysed together rather than treated as three independent documents.

This is particularly valuable in:

• finance
• procurement
• insurance
• logistics
• healthcare
• telecommunications
• utilities

8. Human-in-the-Loop Becomes More Important — Not Less

One misconception about AI automation is that the objective is to eliminate humans from every process.

For many enterprise processes, the better model is:

AI executes the routine 80–95%

Human handles exceptions, risk and accountability

For example:

AI approves automatically when confidence and business rules are satisfied.

AI escalates when confidence is low or the transaction exceeds a defined threshold.

This creates a human-in-the-loop operating model.

The key question becomes not:

"How do we remove humans?"

but:

"Where should humans remain in control?"

This is particularly important for financial transactions, HR decisions, cybersecurity and other sensitive processes.

9. Process Mining + AI + RPA

Process mining is becoming an important foundation for automation.

Instead of asking employees:

"How does this process work?"

organisations can analyse actual system event data.

Process mining can identify:

• bottlenecks
• repetitive activities
• unnecessary approvals
• process variations
• rework
• exceptions
• manual intervention
• automation opportunities

AI can then analyse the discovered process and recommend potential improvements.

This creates a powerful cycle:

Discover → Analyse → Redesign → Automate → Measure → Improve

This is much more sophisticated than simply creating individual RPA bots.

10. RPA Governance Becomes a Strategic Requirement

As automation becomes more intelligent, governance becomes more important.

Enterprises need to know:

• Which AI agents exist?
• What systems can they access?
• What data can they read?
• What transactions can they execute?
• Who owns the process?
• What decisions can they make autonomously?
• When must they ask for human approval?
• How are their decisions logged?
• How are failures detected?

A mature enterprise automation model therefore needs:

Identity & Access Management

AI governance

Process governance

Auditability

Security

Human oversight

Monitoring

Without these controls, scaling automation can also scale operational risk.

11. RPA Is Moving Toward "Automation Fabric"

The long-term architecture is increasingly becoming an enterprise automation fabric rather than an RPA platform.

Imagine an architecture consisting of:

BUSINESS OUTCOME │ ▼ AI AGENT LAYER │ ▼ PROCESS ORCHESTRATION │ ┌──────────────┼──────────────┐ ▼ ▼ ▼ RPA APIs AI/ML │ │ │ └──────────────┼──────────────┘ ▼ ENTERPRISE SYSTEMS │ ┌──────────────┼──────────────┐ ▼ ▼ ▼ SAP CRM LEGACY

This architecture allows different technologies to perform the jobs for which they are best suited.

RPA should not disappear.

Instead, it becomes one component of a much broader automation ecosystem.

12. The Rise of "AI-First Automation"

The strategic question is also changing.

Five years ago, companies asked:

Where can we deploy an RPA bot?

Today the question is increasingly:

Which business processes should be redesigned around AI, automation and orchestration?

This is a fundamental shift.

Automation is moving from an IT efficiency initiative toward an enterprise transformation capability.

Gartner's 2026 research on agentic AI also points toward specialised, domain-specific agents as an important route to measurable business value rather than relying solely on generic agents.

What Does This Mean for SAP Transformation?

For organisations running complex SAP environments, the opportunity is particularly interesting.

Consider a future SAP automation architecture:

SAP S/4HANA

↕

Enterprise Integration / APIs

↕

Process Orchestration

↕

AI Agents

↕

RPA / Computer Use

↕

Legacy applications

This could automate processes such as:

• supplier onboarding
• invoice processing
• purchase-to-pay
• order-to-cash
• master-data management
• user access requests
• SAP incident resolution
• testing
• reconciliation
• reporting
• data migration activities

The important architectural principle is:

Do not use RPA to compensate for bad architecture.

Where a clean API exists, use the API.

Where a process should be redesigned, redesign it.

Where AI is required for interpretation, use an AI capability.

Where a legacy application cannot easily be integrated, RPA may remain the appropriate bridge.

The 2026 RPA Technology Stack

I would describe the emerging stack as five layers:

Layer Main capability
AI Agents Reasoning, decisions, autonomous task execution
Generative AI Understanding language and unstructured information
Process Orchestration Coordinating people, systems and automation
RPA / Computer Use Interacting with applications
APIs & Integration Reliable system-to-system transactions

And surrounding all five:

Security + Governance + Observability + Human Oversight

The Biggest Change: From Bots to Digital Workers

The concept of the digital worker is becoming more realistic.

A traditional bot performs a predefined task.

A digital worker could potentially:

• understand instructions
• access multiple systems
• interpret documents
• make bounded decisions
• execute transactions
• communicate with users
• monitor outcomes
• escalate exceptions

This is why the boundary between:

RPA

AI

workflow automation

integration

and digital assistants

is increasingly disappearing.

Five Questions CIOs Should Ask

Before launching another RPA program, CIOs should ask:

1. Are we automating tasks or redesigning processes?

2. Could AI agents handle the cognitive part of the process?

3. Should we use an API instead of RPA?

4. Where must humans remain accountable?

5. Can the automation architecture scale beyond individual bots?

These questions shift the conversation from RPA implementation to enterprise automation strategy.

Conclusion

The future of RPA is not simply more robots.

It is the convergence of:

RPA + Generative AI + AI Agents + Process Mining + Orchestration + APIs + Human Oversight

Traditional RPA remains valuable because enterprises still need reliable automation of structured UI-based tasks.

But the strategic direction is clear: automation is moving upward from individual tasks toward end-to-end, intelligent and increasingly autonomous business processes.

The winners in this next phase will not necessarily be organisations with the largest number of bots.

They will be organisations that understand which processes should be automated, which should be redesigned, where AI should make decisions, where RPA should execute actions, and where humans must remain in control.

A useful transformation formula

Traditional RPA

Rules → Bot → Task

Intelligent Automation

Rules + AI → Automation → Process

Agentic Automation

Business Intent → AI Agent → Orchestration → Systems → Outcome

That is the real evolution of RPA in 2026.