Skip to main content

Let us submit AI drafts to the approval queue via API (BYOK / bring your own AI)

Right now the API can do two things with a reply: send it, or post an internal note. There is nothing in between.

That means if you want to use your own AI to draft replies, your only options are:

  1. Let it send directly to the customer, which nobody sensible wants unsupervised, or

  2. Post the draft as a private note and have an agent read it and copy-paste it into the reply box

Option 2 works, but it throws away the best thing Commslayer already built. The "AI needs approval" queue is genuinely good. One click, the agent reviews, it goes. We just cannot put anything into it unless it came from the built-in agent.

The request: an API endpoint to create a draft reply in the pending-approval state, exactly where the built-in AI's drafts land. Same queue, same review, same one-click approve. Just a different author.

Why this helps everyone, not just us:

  • Teams with their own fine-tuned model or brand voice can use it without giving up the human gate

  • Teams with no AI at all can use it for templated or rules-based drafts

  • It is opt-in, changes nothing for anyone who does not want it

  • It is the safest possible way to add automation, because a human still approves every message

Why we are asking: we run a children's clothing brand and tone is most of our product. Our customers are mums, often anxious about an order arriving for a birthday. The built-in agent is good at triage and genuinely reliable at handing over when it should. What it cannot do is sound like us, and we cannot teach it to, because we cannot bring our own model. So our best replies currently have to be pasted in by hand, which is slower than doing nothing at all.

The approval queue already exists. We would just like to be allowed to put something in it.

Status: Completed4 comments

Comments4

  • Karri Koivuniemi

    Team•

    Sep 22

    We’ll implement this 👍

    • An Anonymous User

      •

      Sep 23

      Thats amazing Karri! Any idea on approximate time frame? We can help test the feature at any time if that is helpful and happy to jump on a call if would help you in any way before implementing.

    • Jade Andy

      •

      Sep 23

      Thats amazing Karri! Any idea on approximate time frame? We can help test the feature at any time if that is helpful and happy to jump on a call if would help you in any way before implementing.

      • Karri Koivuniemi

        Team•

        Sep 25

        Done!