Automating half my support team's work moved the bottleneck onto me

I lead a support team of 14 at a B2B SaaS. Over the last eight months I built out an agent pipeline (Fin for tier-1 triage, a homegrown RAG layer over our Zendesk macros, GPT drafting for tier-2 replies). Deflection went from 19% to 51%. CSAT held at 4.6. On paper, a win. What nobody warned me about: the work that's left is the work that breaks people. My agents handle the 'where's my invoice' tickets. My humans now spend their day on angry enterprise escalations, billing disputes, and edge cases the bot punted. Two of my best reps quit in Q1. Exit interviews both said the same thing, the job stopped having easy wins. I also became the bottleneck. Every agent prompt change, every macro update, every edge case triage rule routes through me. I'm doing four jobs: manager, prompt engineer, QA, and escalation handler. My calendar is unrecognizable. The productivity number is real. The org chart we need to support it does not exist yet. I don't think mine is the only team in this spot.

▲ 40·thabo_mokoena·4w

11 comments

0yara_najjar·3w

Same pattern here after we shipped a triage bot for our 12 person support team. Ticket volume halved but every novel edge case now routes to me for prompt tweaks or tool fixes, and I'm the single point of failure for the queue. Curious if anyone has actually pushed prompt ownership back onto the support leads instead of keeping it on engineering.

0aminataDiallo·3w

Saw this on the design side too. We shipped an AI ticket triager last quarter and now half my week is reviewing the auto-generated UX bug reports it files into Linear, most of which are dupes or misread screenshots. The throughput gain on their side just became my backlog.

0thomas_weber·3w

Same shape on our side. I shipped a triage bot for a 6 person support team and now every edge case it can't handle lands in my queue, plus I'm the one explaining to support leads why the confidence threshold dropped this week. Did your scope officially change or are you just absorbing it?

0sofiaAlves·3w

Same pattern hit me last quarter. I used to write 4 long-form pieces a week, now the agency expects 12 because "drafting is solved," and every hour I save on first drafts gets eaten by reviewing AI output that's 80% there but wrong in subtle ways only I catch.

0joaqd·3w

Same pattern hit us when I shipped triage automation for a 12-person CX team. Now I'm the SPOF for prompt tweaks, edge cases, and the weekly "why did it close this ticket" review, and there's no clean way to hand that back without retraining someone on the whole stack.

0joaqd·3w

Saw a smaller version of this when I let an AI grader handle first-pass feedback on essays for my two sections of 28. Throughput on returning drafts tripled, but now I'm the choke point on the kids who push back on the feedback, and those conversations are the actual teaching. Curious whether your support team is seeing the residual cases get harder on average, or just more concentrated.

0sam_okafor·3w

Same pattern here after I wired up an LLM triage layer for our 4-person support queue. Tickets resolved faster, but every escalation now lands in my DMs with three paragraphs of context I have to actually read, and the volume is higher because the easy filter is gone. Did you end up hiring a senior IC to absorb the new top-of-funnel, or restructure how escalations get routed?

0omarKhaled·3w

Same thing happened on our side after I shipped a triage bot for a 6 person support team. Now every weird edge case escalates to me directly, and I'm basically on call for the automation I built. Did you end up training anyone on the team to debug the prompts, or is it all still funneling through you?

0MiaJ·2w

Watched the same thing happen at a logistics startup last year, except their bottleneck shifted to QA on the bot's edge cases, and the senior agent who used to handle escalations quit within four months because her job became reviewing transcripts instead of talking to customers. Worth asking whether the work you absorbed is actually leadership work or just the messy residue nobody scoped before automating.

0karim_r·2w

Same trap on our side, the AI ticket triage shipped fine but every macro response needed a UI state I had not designed yet. I went from two Figma reviews a week to seven, and the support lead is now blocked on me for empty states and error toasts.

0arjunsharma_ml·2w

Moved the bottleneck onto me" is the part everyone glosses over. You didn't automate half the work, you converted ticket-handling labor into reviewing-LLM-output labor and concentrated it on one person who can't be parallelized. That's not a 50% gain, that's a single point of failure with a Zapier subscription.