By Piet Baudoin · 14 November 2026
At the 60 accounting firms in SRA’s culture survey, people look into why something went wrong more often, while timely and complete information sharing is declining. Also: AI that starts working during the meeting, and memory for customer conversations.
Who tells everyone else what changed?
“How can accounting firms grow without connection, quality and job satisfaction coming under pressure?” (translated from Dutch) With that question SRA, a Dutch association of accounting firms, presented its culture survey on 11 November. At the 60 firms that took part, collaboration scores an average of 7.9. Still, SRA writes: “Employees are experiencing a greater workload.” (translated from Dutch)
The factsheet that comes with the survey counts 2,206 completed questionnaires, collected from 1 September 2025 through 1 November 2026, and puts two short lists side by side. Among the biggest risers since the previous survey is carrying out root-cause analyses. Among the biggest fallers are timely and complete information sharing and reviewing assignments. So people look into why something went wrong more often, while passing on information scores lower than in the previous survey.
Do you recognize the situation where someone resolves an exception while everyone else still works from the old agreement?
Source: SRA, 11 November 2026 · SRA culture surveys factsheet, 11 November 2026
Can AI start work during the meeting?
On 12 November, Anthropic began rolling out Claude Meetings, a beta for Team and Enterprise plans, for now only in Google Meet. Claude joins the call as a visible participant and can immediately draft a plan that the participants edit. Anthropic sells the service itself.
One detail about access to participants’ connected tools: Claude asks the first time, “and a yes applies going forward”. For spoken requests, though, Claude currently asks each time who is speaking. So giving permission and being recognized are two separate steps. Anthropic also asks you to check beforehand that everyone on the call is OK with being transcribed.
Source: Anthropic, 12 November 2026
Does a returning customer have to explain everything again?
On 11 November, ElevenLabs announced Memory for its conversational agents. The feature is intended to use earlier conversations after recognizing an authenticated customer. The company sells this software; this is its announcement, not an independent field test.
ElevenLabs writes: “the CRM remains the primary customer record”. The extra memory keeps conversational details that had no fixed field, such as when someone can take a call. So the announced memory adds to the customer record.
Source: ElevenLabs, 11 November 2026
Voices
Developer Zoop, active on Tweakers, a Dutch technology site, since 2004, described on 22 October, under an article about an AI agent from Anthropic, how he works: he pastes a detailed report from a customer into an agent, and puts the plan that comes out next to the customer’s question. Asked what his added value still is, he answered the next morning:
Read and check everything. If the bot builds something that isn't right, I'm responsible, not the bot. So my added value is indispensable. (translated from Dutch)
Zoop · 23 October 2026
It is his own experience; no one in that thread confirmed it.
Where are we headed?
Claude can already draft a plan during a conversation that the participants edit right away. I expect that within three years a team will run through a busy working week digitally before it begins. For that, AI first has to learn from earlier assignments how the work actually goes, and read images of the work itself next to the agreements on paper. And the team has to agree on which information it may use together. Then AI shows during the meeting where the schedule gets stuck if a delivery doesn’t arrive. Everyone sees the same consequences, and the people at the table decide together which agreement changes. The week then starts with a plan that already includes that decision.
What does this mean for your business?
All three stories are about what is kept after a conversation or a correction, and by whom. Bombos starts with one task, often the shared inbox: an AI agent prepares replies to recurring customer questions, you check them, and a correction you give is carried over to the next time, so you don’t have to explain the same way of working over and over. After that, your team teaches Bombos the next task itself. The free scan shows where it could start for you.
Sources
Every source was opened on 14 November 2026; every quote appears in it word for word, quotes from Dutch sources in our translation.
SRA: Kwaliteitsgerichte cultuur groeit door meten en bespreken (11 November 2026) · SRA: Culture surveys factsheet (11 November 2026) · Anthropic: Introducing Claude Meetings (12 November 2026) · ElevenLabs: Everything announced at Summit NYC (11 November 2026) · Zoop on Tweakers, 22 October and 23 October 2026
Bombos team