Orchestration. Every sample manager knows that they need it, but everyone has a different idea of what it means. In summer 2026 Cenevo held a focus group with nine Mosaic sample managers to discuss the topic of Orchestration, Scheduling, and Workflow with a particular emphasis on current pain points.
The Design-Make-Test-Analyze (DMTA) cycle featured heavily in the conversation. This was no surprise, given it is the fundamental process of life sciences discovery research, and optimizing the cycle time is key to delivering new medicines to patients more quickly.
The DMTA cycle brings together staff from multiple departments and disciplines who are invariably using multiple different software systems. Orchestrating the entire DMTA cycle requires coordination of the physical processes, communication between departments, and integration of the software systems.
There is a lead time to prepare reagents and cell lines for an assay, meaning that advance notice of the number of expected samples is helpful. However, whilst research scientists have a plan for the number of new molecules they intend to make and submit for testing, some will fail QC, and so the actual number of samples sent to the assay may differ from the original plan.
The focus group discussion supports the finding of our 2026 Lab Operations Report that 55% of life sciences professionals cite lack of integration between systems as their biggest barrier to making effective use of lab data. The DMTA cycle makes that challenge acutely visible.
There was consensus amongst the focus group attendees that ease of integration of software systems is a key requirement of orchestrating the DMTA cycle. Whilst it is superficially attractive to have a single software system managing the entire process, in the real-world different departments will each select the software tools which are optimal for their part of the process. The IT department may also be wary of risk of having a single huge monolithic system which attempts to do everything.
Therefore, it is important that software systems have high-quality programming APIs to enable integrations. However, having APIs is not sufficient — the integrations still need to be built. The recent explosion in the availability and quality of AI coding tools is making it easier than ever to create integrations between systems, and allows non-programmers to create integrations using low-code and no-code tooling. But there is still a need to provide a platform to host, test, and run the integrations which complies with corporate IT policies. I expect that we will increasingly see AI coding and hosting platforms being built-in to the software you already use and trust, such as Cenevo’s Mosaic and Labguru.
Equally important is for programming APIs to have high-quality documentation which can be understood by both humans and AI. REST APIs with OpenAPI Specification (OAS) documentation (in JSON or YAML) are one part of the solution. Self-describing Model Context Protocol (MCP) APIs are another part of the solution.
One of the many jobs for sample managers is to monitor the stored stocks and identify when samples are nearing depletion. The actions to take on depletion can be highly variable, and may include solubilizing neat stock, transferring stock between research sites, notifying the research scientist to consider synthesizing new samples, etc. This is a problem ideally suited to agentic AI tooling. Sample managers can describe their re-stocking requirements in natural language, and have the AI convert that into program script code to implement the rules. Indeed, our 2026 research shows 27% of labs are already exploring agentic AI.
Orchestration within the realm of sample management is as important as orchestration across the entire DMTA cycle. This is the area where the terms “orchestration”, “scheduling” and “workflow” have the most overlap, with no universally agreed distinction.
The rules for choosing between the possible samples to fulfil an order was considered to be orchestration by some of the focus group attendees. Other attendees said that the rules for prioritizing orders when there are too many to complete within the day was an important constituent of orchestration.
Interestingly, there was virtually no discussion in the focus group of orchestrating and scheduling multiple devices within an automation workcell. However, I believe this should not be taken to mean that this is not an important topic. It perhaps instead indicates that this is largely a solved problem, and is not a significant pain point.
Hand-in-hand with orchestrating a process, is the desire to monitor and improve both the orchestration and the process itself. For this it is critical to have detailed data in a form which can be easily analyzed. As one focus group attendee observed, you would draw very different conclusions from the data that an order took 2 weeks elapsed time to process, compared to data saying that the order was only actively being processed for 2 hours, and spent the rest of the elapsed time idle.
In conclusion, it was clear that effective orchestration needs to cover people, processes, and software. All of these must work together, and different areas will be of varying importance at different times as organizations work to eliminate pain points.
However, we did not come up with precise definitions to distinguish the terms “orchestration”, “scheduling” and “workflow”. And in the end, perhaps that doesn’t matter. As Shakespeare wrote, a rose by any other name would smell as sweet.