Skip to content
AYBIZA

Your first support hire is a tier-one decision.

You are answering questions at night that your own documentation already answers. Hiring solves the symptom.

There is a stage every software company reaches where the founder is still the support team. Questions land in a shared inbox and a channel, and they get answered late, by whoever has the least broken concentration.

The tell that you have reached the end of that stage is not volume. It is repetition. You start recognising the questions. You start copying your previous answer. And somewhere in there you realise that almost every answer you send already exists, written down, in your own documentation.

That is the moment the hiring conversation starts. It is usually the wrong conversation.

What you would actually be hiring for

Sort the last month of questions into two piles.

The first pile is tier one: how do I do this, where is that setting, why did this fail, what does this error mean, does the plan include that. The answers exist. They are in your docs, your changelog, or a thread from a while back. They need retrieving and phrasing, not deciding.

The second pile is everything else: the bug nobody has seen, the customer deciding whether to renew, the integration that needs an engineer.

A support hire absorbs both piles, which is why hiring feels like it works. But you are paying a salary primarily to have your existing documentation read back to people, and the second pile — the one that actually needs a human — was never the bottleneck.

Why the bot you might build will not fix it

The reflex is to point a model at the docs and put a bot on the site. Most teams at this stage have already tried it, and most of those bots are sitting unused.

The reason is rarely answer quality. It is that the bot lives outside the work. It answers on a page nobody on the team visits, it has no permissions of its own, and it cannot do anything with what it learns. When a customer question needs a person, the bot cannot bring one — it can only stop.

So the team keeps working the way it always did, and the bot becomes a thing that exists.

What tier one looks like handled properly

An agent takes the first pass on the widget and on the inbound line. It answers from your knowledge, against the customer’s actual record, and it says so plainly when it does not know.

When the question belongs to a person, it escalates into the channel where your team already works, with the conversation and the record attached. Not a notification pointing at another tool. The thing itself, in the room, ready to be picked up.

Your qualified leads get the same treatment. A question from someone evaluating you gets an answer while they are still evaluating, instead of six hours later when they have moved on.

And because the agent is a member of the workspace rather than a service on a page, your team can mention it, correct it, and hand work to it the way they would with anyone else.

The actual decision

The question is not whether to hire someone. You will, eventually, and they will be valuable.

The question is what you want them working on when they arrive. If tier one is already handled, your first support hire spends their time on retention, on the hard cases, and on making the product need less support. If it is not, they spend it reading your documentation aloud, and you will be having this conversation again about the second hire.

Frequently asked questions.

Should I hire a support person or automate tier one first?

Handle tier one first when the bulk of your volume is questions your documentation already answers. Hiring absorbs that work rather than removing it, and the new hire spends their time retrieving answers instead of on retention and hard cases. Automating first changes what the eventual hire is for.

What counts as a tier-one support question?

A question whose answer already exists and needs retrieving and phrasing rather than deciding — how to do something, where a setting lives, what an error means, what a plan includes. Tier two is anything requiring judgement, investigation, or a decision.

I already built a docs chatbot and nobody uses it. Why?

Usually because it lives outside the work. It answers on a page the team does not visit, holds no permissions, and cannot act on or escalate what it learns. Adoption depends on the agent being a member of the workspace, not on the quality of its answers.

How does an agent hand a conversation to a person?

It escalates into the channel where the team already works, with the conversation and the customer record attached, so whoever picks it up starts in context.

Can the agent answer from our existing documentation?

Yes. Agents answer from your knowledge base alongside the customer's record, so the response reflects both what your documentation says and who is asking.

Does this only apply to support, or to sales questions too?

Both. Inbound qualification has the same shape — a first pass that answers immediately and brings a person in when the conversation is worth their attention.