Self serve sales and when to add humans

Self serve is not the absence of sales. It is sales performed by the product, at a marginal cost of roughly zero, on a schedule the buyer controls. The pricing page is the pitch. The trial is the demo. Checkout is the close.

Once you see it that way, the real question changes. It stops being “when do we start selling” and becomes “which specific job can a human do here that the product cannot”, and that question has a much smaller answer than most teams want.

What the product has to do unaided

Before you hire anyone, the self serve path has to survive a buyer who wants to spend money and has a procurement department.

That means checkout handles more than a card. It needs to calculate VAT and sales tax correctly for the buyer’s jurisdiction, add and remove seats mid cycle with sane proration, produce an invoice with a company name and address rather than a receipt with a personal email, and offer a purchase order or bank transfer path above some threshold. Most self serve products fail at that last one and lose deals silently, because the buyer never writes in to complain, they just stop.

The upgrade path needs to be reachable from inside the product at the moment of friction, not only from a pricing page. When a user hits a limit, the limit screen is the highest converting surface you own. Figma’s editor seat prompt and Calendly’s connected calendar limit both do this well: the wall appears exactly where the value was.

You also need a public security page. Not because it converts directly, but because it unblocks the IT reviewer who has been handed your product by a champion, and because visits to it are one of the highest value signals you will collect.

Which account signals justify a human?

Account level ones. Individual behaviour is noise in B2B, because the person clicking is often not the person buying.

The signals that hold up across companies I have looked at are seat count crossing a threshold you set, users appearing on two or more email domains inside a single workspace, a second or third admin being invited, API keys being generated, SSO or workspace settings being opened, and visits to security, compliance or procurement pages on your marketing site.

Signal	What it implies	Strength alone	Best pairing
Seat count crosses threshold	Team adoption underway	Medium	Second admin invited
Two or more email domains in account	Multi entity or agency use, contract complexity	Medium high	Invoice request
Second admin invited	Governance is being set up, so someone is standardising	High	Seat growth in 14 days
API key generated	Technical integration, higher switching cost	Medium	Usage growth
SSO settings opened	IT is involved, enterprise plan in scope	Very high	Security page visit
Security or procurement page visit	Formal review starting	High	SSO settings opened
Single user upgrading to paid	Individual value, not a deal	Low	Nothing, leave it alone

The last row is the discipline test. A solo user upgrading to a 15 dollar plan is not a sales opportunity, and a rep who contacts them is burning a sixty dollar touch on a deal the product already closed. Build the threshold into the CRM so reps cannot see those accounts at all. Visibility is the control, not policy.

Slack ran the clearest public version of this for years: the free workspace grew on its own, and humans appeared when a company wanted org wide deployment, compliance controls and a single bill. Notion and Figma draw a similar line. In each case, the human enters at consolidation, not at first payment.

How does a sales assist rep actually spend their week?

Not on discovery. The product did that, and the usage data tells them more than a discovery call would.

A sales assist rep’s week is mostly seat consolidation conversations (you have four separate paid workspaces in this company, here is one contract), security questionnaires, procurement paperwork, and rollout planning with a champion who wants to bring in three more teams but does not have the internal standing to do it alone.

That is a real job, and it needs different hiring than a full cycle AE. You want someone comfortable reading product usage dashboards, patient with paperwork, and unbothered by short cycles with small individual values. A hunter from a six figure ACV background will be miserable and will start trying to run discovery calls on accounts that do not need them.

Quota structure matters as much as hiring. Comp on total revenue in assigned accounts and you pay reps for the product’s work. Comp on incremental revenue above a self serve baseline, or a fixed bounty per successfully expanded or consolidated account, and you pay for what they actually added. The baseline is the hard part, and the holdout gives it to you.

Designing the holdout that proves incrementality

Here is the test almost nobody runs, and the reason most sales assist programs cannot defend themselves in a downturn.

Expect an uncomfortable result the first time. Common outcomes are a strong lift in the top quartile of qualifying accounts and no measurable lift at all in the bottom half, which is exactly the finding you want, because it tells you to raise the threshold rather than fire anyone.

The holdout also resolves the argument that never resolves otherwise: marketing says the funnel produced it, sales says the rep closed it, and attribution reports agree with whoever built them. A randomised holdout is the only evidence that survives that meeting. The same logic underpins the measurement section of the PLG metric set.

Handoff rules that stop sales taxing the funnel

Three rules, all enforced in software rather than in a team agreement.

First, an account below the threshold is invisible to sales. Not deprioritised, invisible. CRM visibility rules, not a Slack norm.

Second, outreach has a cooldown. One touch sequence per account per quarter, maximum three messages. Self serve users react badly to being hunted, and the churn cost of an annoyed admin is larger than the expected value of a fourth email.

Third, the rep cannot block the self serve path. If a rep is engaged and the champion decides to just buy 20 seats on a card, that has to work. Teams that route qualified accounts into a “contact sales” wall convert worse overall, because they have replaced a working purchase mechanism with a calendar link and a two day delay.

There is a genuine tradeoff here worth naming. Enforcing invisibility means you will miss some real opportunities in the sub threshold band, and a rep will eventually show you an account they could have grown if only they had seen it. That anecdote is true and it is not an argument, because it is available in every system regardless of where the line sits. Move the threshold based on the holdout, not based on the anecdote.

For the wider strategic framing of when each motion applies, self serve vs sales assisted sets the two side by side, and choosing a SaaS sales model by ACV is the sizing question underneath it.

When self serve is the wrong motion entirely

Sometimes the answer is that you should not be self serve at all, and it is worth checking before you invest another quarter.

Self serve struggles when a single user cannot get value without their organisation doing something first: connecting a data warehouse, signing a data processing agreement, or getting IT to install an agent. It struggles when the product changes a regulated process, because nobody adopts a compliance workflow from a credit card purchase. And it struggles when the ACV required to cover your cost to serve exceeds roughly 15 thousand a year, because at that price buyers expect a human and procurement demands one.

If two of those three describe you, adding sales assist is treating a symptom. Read product led growth vs sales led growth and the longer form PLG versus sales led comparison for B2B before you hire, and look at how comparable companies structured it in the sales process teardowns.

What to do in the next 30 days

Write the threshold rule as a single sentence a CRM can execute. Then count how many accounts crossed it in the last four quarters. If the answer is under 60 a quarter, you do not have a rep’s worth of work, and the honest move is to improve the upgrade path rather than hire.

If the volume is there, launch with the holdout from day one. Retrofitting a holdout after six months of untracked outreach is close to impossible, because your reps have already contaminated the comparison group. Start at the SaaS sales overview if you are still choosing a motion, and the startup sales strategy playbook if the first hire is imminent. The gating decision itself is covered in lesson 2 on choosing your go to market motion.