A representative German survey just measured the demand side of grid flexibility, and half of it does not know it exists.
Bitkom asked German households in August whether they had heard of bidirectional charging. Forty-six percent said yes. Forty-five percent said they had never heard of it at all. Those two numbers describe the same object: a car with a usable battery sitting in a driveway, capable of feeding power back into a house or a grid, owned by people who split almost evenly on whether that capability exists.
This matters because of what the rest of the survey says. Seventy-seven percent would feel safer during a power outage with an electric car acting as a storage unit. Sixty-four percent find the idea practical. Fifty-two percent see potential for secure energy supply, and forty-four percent want stronger political support for it. Once people understand the function, they like it. The gap is not resistance. The gap is that nobody told them.
Autonomous energy operations assume a supply that has not been informed
Roughly a third of energy operations are already described as fully autonomous, with an industry target of half by 2030. The architecture behind that target depends on aggregating distributed flexibility at the grid edge: heat pumps, home batteries, and above all vehicle batteries that can absorb a surplus at noon and release it at seven in the evening. Software agents are supposed to orchestrate this. The engineering is the interesting part, and it is also the part that gets all the attention.
The Bitkom number puts a different constraint in front of that engineering. An agent can only dispatch flexibility it has been granted. Granting it requires an owner who signs something, accepts a tariff, or clicks through a consent screen. Half of the owners in Germany do not yet have a mental model for what they would be consenting to. That is not a technology problem, and it will not be solved by a better optimiser.
There is a useful way to read the 46 versus 45 split. It is not a slow adoption curve that will fill in on its own. It is a distribution problem, and distribution problems have channels.
The aggregator is turning out to be the channel everywhere
In April, Thüga signed a framework agreement with Deutsche Telekom and T-Systems for sovereign enterprise AI, procured centrally for the municipal energy and water utilities in its network. Thüga covers around 100 Stadtwerke, the largest such network in Germany, and the stated logic of the deal is that no individual utility should have to carry the integration itself. Data processing stays inside European jurisdiction.
If this architecture holds, you would expect the same shape to appear in adjacent fragmented industries. It does. In construction, the reachable aggregator is the building materials distributor, which is why the Higharc and US LBM arrangement from June is worth more attention than most construction software news. In the trades, it is the chamber and the Kreishandwerkerschaft. Three industries, three sets of thousands of small buyers, one repeating answer: you reach them through the intermediary they already transact with weekly.
Bidirectional charging follows the same logic and nobody has claimed the position yet. The household will not be educated by a vehicle manufacturer's press release. It will be educated by whoever already sends it a bill: the Stadtwerk. Thüga's hundred utilities are the same hundred utilities that would need to explain, tariff, and settle a vehicle-to-grid arrangement with the households in their supply area. The AI procurement and the flexibility question are arriving at the same desk, and right now they are being handled as separate topics by separate people.
The counter-evidence sits in a KfW report
KfW Research published Fokus Nr. 554 in late July, and its central finding is uncomfortable for anyone selling specialised industrial AI. What the German Mittelstand actually adopts are technologies that are easy to handle with broad application areas. Not deep vertical systems. Not orchestration layers. The same report puts the construction sector last on every single AI technology measured, which is a headline everyone will quote, and buries the more important point underneath it: buyers are moving toward low-threshold tools, not sophisticated ones.
Applied here, that cuts both ways. It argues against pitching a household or a small utility on grid-edge agent orchestration. It argues for the version of the offer that a customer can understand in one sentence, which is roughly what the Bitkom survey already found people responding to: your car keeps the lights on when the power goes out.
The practical move is to stop treating awareness as marketing overhead and start treating it as the actual product surface, because in a market where half the owners of an asset do not know what the asset does, the person who explains it owns the relationship that any later optimisation gets sold through.
Two data points from the same summer, then. Half of German households have never heard of the capability their next car will ship with, and a hundred municipal utilities just bought AI centrally through a single framework. Those facts are currently unconnected. Whether they stay that way is the open question for the next eighteen months, and the answer will be decided by people writing customer communication, not by people writing dispatch algorithms.