The single biggest predictor of an accident in IFR conditions is whether there's one pilot in the cockpit or two. This is not subtle. The Part 91 single-pilot turbine accident rate is meaningfully higher than the Part 135 two-pilot rate flying the same airplanes in the same conditions. The pilots are equally capable. The airplanes are equally airworthy. The difference is the cockpit.
What kills you in single-pilot IFR isn't a single failure. It's the workload curve, and where on it you happen to be when something else goes wrong.
The curve
Cruise flight is easy. The airplane is trimmed, the autopilot is doing the work, the workload is low. You can talk on the radio, eat your sandwich, look at the chart, plan the approach, brief yourself on the missed. Single-pilot or two-pilot, doesn't matter much. The cockpit isn't busy.
The approach is different. In a typical IFR approach you're doing all of the following more or less simultaneously. Setting up the FMS for the approach. Briefing yourself on the procedure. Configuring the airplane for landing, which means flaps, gear, power changes, prop adjustments depending on the airplane. Communicating with approach control. Monitoring weather, especially if it's marginal. Watching for traffic. Configuring the navigation to the approach. Following the published procedure, which has altitude crossings and a stepdown profile. Adjusting the autopilot or hand-flying. Watching the airspeed bleed off as you slow. Anticipating the missed approach in case you need it.
This is a lot. In a two-pilot cockpit you split it. The pilot flying focuses on the airplane. The pilot monitoring handles radios, configuration, briefings, and looks out the window. The workload divides into two roughly manageable streams.
In a single-pilot cockpit, one person does all of it. The workload isn't divided. It's serialized, which means while you're doing one thing, you're not doing the others. This is fine when nothing surprises you. When something does, the surprise lands in a cockpit where everything is already full.
What goes wrong
The NTSB chains for single-pilot IFR accidents read the same way over and over.
The pilot is on approach. Workload is high. ATC issues an unexpected instruction that breaks the rhythm. The pilot acknowledges but doesn't quite process. The airplane is configured incorrectly for the new clearance, or descends through an altitude that wasn't actually cleared, or turns onto a heading that takes it toward terrain. The pilot is now behind the airplane. The cascade starts.
Or the pilot is in the missed approach. The transition from descending to climbing, from landing config to clean, from approach frequency to departure frequency, from one mental model to another. The workload spikes. Something gets dropped. Sometimes that something is critical.
Or the pilot is in cruise, encounters something unexpected like icing or a system anomaly, and the diagnosis steals attention from the basic flying. The airplane drifts off altitude. The autopilot is doing its job but the pilot isn't checking. By the time anyone notices, the situation has compounded.
These aren't pilots who didn't know how to fly. These are pilots who ran out of bandwidth at the wrong moment.
What helps
Two pilots help. Obviously. But most Part 91 operations are single-pilot, and that isn't changing.
A capable autopilot helps. Hand-flying an IMC approach while doing everything else is the highest-workload version of this problem. A well-coupled autopilot moves the basic flying off your plate.
Standard operating procedures help. Flow patterns help. Checklist discipline helps. Anything that turns a cognitive task into a habitual one frees up bandwidth for the things that actually need attention.
What's been missing is anything that monitors the cockpit and tells you when something doesn't add up. The autopilot will fly the heading you set. It won't tell you that the heading you set doesn't match what ATC just told you. The altimeter will show your altitude. It won't tell you that you're descending through an altitude you shouldn't be.
This is the workload offload that hasn't existed in the GA cockpit. The pieces are now buildable. An advisory layer that listens to ATC, reads aircraft state, and tells you when they don't match. Not autonomy. Not envelope protection. Just the second set of eyes that the right-seat pilot would have provided, if there were a right-seat pilot.
For single-pilot IFR, this is the missing piece. The question isn't whether the technology is ready. It is. The question is who builds the version that works without adding more workload than it removes.