Is my RPA investment now worthless?

No. RPA remains the right choice for rule-based, structured, high-volume work. It shifts from a strategic to an execution layer beneath AI agents.

How do I recognise real agentic AI instead of agent washing?

Real agents set goals autonomously, choose tools and learn from feedback. RPA bots and chatbots only execute predefined scripts. Ask the vendor specifically.

What does revDSG mean for AI agents in Switzerland?

Personal data must be processed lawfully. Choose hosting in Switzerland or with guaranteed EU/CH data residency, define roles and minimise data.

Can I combine RPA and AI agents?

Yes, that is often the best path. RPA handles structured execution, while the agent orchestrates and decides on exceptions. Existing automations remain usable.

AI Insights

RPA or AI agents? What Sanofi's shift means for your automation

Chris Jon Graf · AI Strategist & CEOPublished on 12 October 2026

Download PDF Briefing

In short

Sanofi is gradually replacing classic RPA bots with autonomous AI agents. For your company, that doesn't make RPA worthless – it makes RPA the execution layer beneath judgment-based agents. Use the matrix: rule-based work remains RPA; unstructured judgment calls belong to the agent.

The short answer: your RPA investment is not worthless. Sanofi shows how RPA shifts from strategic layer to infrastructure layer. The real step is the move from assistants to autonomous AI agents.

The Sanofi case: from RPA to agentic digital labor

Sanofi, the French pharmaceutical group with around 75,000 employees, built its internal assistant Concierge on AWS Bedrock and Snowflake. Around 60,000 employees use it – roughly 80% of the workforce. Concierge is designed to resolve 80% of IT support requests autonomously. The goal: around EUR 10 million in annual savings. Sanofi is also replacing legacy SaaS in contract management, procurement and CRM workflows rather than merely automating them.

EUR 10 million

annual savings targeted by Sanofi through autonomous resolution of 80% of IT support requests.

RPA vs. AI agents: the decision matrix

In practice, the task decides, not the technology. RPA excels at fixed rules and structured data. An AI agent pays off when judgment, context and multiple systems come together.

Rule-based, structured, high-volume work: RPA remains the first choice
Unstructured data, judgment calls, dynamic one-off tasks: AI agent
System integration & data quality: Prerequisite for both approaches
Example: invoice capture with fixed field structure: RPA/BPA
Example: customer enquiry with context and exceptions: AI agent
Role in architecture: RPA becomes execution layer, agent orchestrates

Where RPA remains the right choice

Invoice capture with fixed fields and high volume
Master data reconciliation between clearly defined systems
Rule-based checks with no room for interpretation

Where an AI agent makes the difference

An AI agent pays off when context, judgment and multiple systems come together: customer enquiries, contract review, dynamic order processing. That is exactly where Sanofi scales.

Beware of agent washing and the abandonment trap

Gartner expects that more than 40% of agentic AI projects will be abandoned by the end of 2027. One reason: many vendors practice “agent washing” – selling RPA bots or chatbots as agents. In fact, Gartner counts only around 130 true agentic AI providers. Why governance belongs before deployment is explored in this Swiss AI podcast on scaling.

Scrutinise your vendor

Many tools sold as “AI agents” today are RPA bots or chatbots with a new label. Ask: Can the system independently set goals, choose tools and learn from feedback – or does it just execute predefined scripts?

What this means for your Swiss organisation

For Swiss SMEs, the rule is clear: if you process customer or employee data, revDSG-compliant handling is mandatory. An AI agent that accesses Abacus, Bexio or Microsoft 365 needs clear role and data concepts. Hosting in Switzerland or with guaranteed EU/CH data residency reduces risk. How mid-sized companies embed AI as a leadership topic is discussed in this Swiss AI podcast episode.

The next step

RPA or AI agent is not an either/or question. The decisive factor is matching the task to the technology. A reliable decision starts with an honest look at your processes.

Frequently asked questions

Is my RPA investment now worthless?: No. RPA remains the right choice for rule-based, structured, high-volume work. It shifts from a strategic to an execution layer beneath AI agents.
Which processes should I tackle first with an AI agent?: Unstructured, context-dependent tasks with exceptions: customer enquiries, contract review, dynamic order processing. That's where the judgment component is high.
How do I recognise real agentic AI instead of agent washing?: Real agents set goals autonomously, choose tools and learn from feedback. RPA bots and chatbots only execute predefined scripts. Ask the vendor specifically.
What does revDSG mean for AI agents in Switzerland?: Personal data must be processed lawfully. Choose hosting in Switzerland or with guaranteed EU/CH data residency, define roles and minimise data.
Can I combine RPA and AI agents?: Yes, that is often the best path. RPA handles structured execution, while the agent orchestrates and decides on exceptions. Existing automations remain usable.

Sources

Would you like to explore this topic for your company?

Check Availability

Mass Adoption Has Tipped: AI Competition Now Focuses on Work Redesign The Walden Moment: General-Purpose Robots Enter Production A Barrel of Intelligence for $0.50: What It Means for Your Budget