Artisan

Outbound sales is one of the few white-collar functions built almost entirely out of repeatable steps. Find companies matching a profile. Find the right person. Verify the email. Research something specific enough to open with. Write. Send. Follow up four times. Handle the reply. Book the meeting.

Every one of those steps has a software vendor attached to it, which is why a typical outbound stack is five or six tools with a junior rep in the middle moving data between them and doing the parts nothing automates.

Artisan's argument is that the person in the middle is the wrong shape for the job. This review covers what consolidating that stack actually fixes, why deliverability is the constraint everyone underestimates, how to read the AI-employee framing, and the specific conditions where Clay or a human rep remains the better answer.

The AI BDR Proposition

• Artisan is an outbound sales platform built around Ava, an autonomous AI business development representative that finds B2B leads, writes personalised outreach, sends it, handles the replies and books meetings.
• Consolidation is the real product. Ava replaces a stack that normally means separate vendors for contact data, enrichment, email deliverability, sequencing, dialling and intent signals.
• Founded in 2023 by Jaspar Carmichael-Jack, San Francisco-based, with a $25 million Series A led by Glade Brook Capital alongside HubSpot Ventures, Y Combinator and Day One Ventures.
• Ava 2.0 is the current generation, positioned for enterprise teams rather than only startups.
• Charging is credit-based with tiers from startup through enterprise — you buy capacity for the agent to work, not a seat for a person.

Consolidation Is A Bigger Claim Than Autonomy

The headline is the autonomous agent. The more defensible value is the stack it collapses, and it is worth separating the two.

A functioning outbound operation typically runs a contact database, an enrichment layer, an email verification and deliverability service, a sequencing platform, an intent-data provider and a dialler. A typical stack pairs Apollo.io for data with Instantly for sending. Six contracts, six renewal dates, six admin surfaces, and integration seams between all of them.

Those seams are where outbound actually breaks. Data goes stale between the database and the sequencer. A contact gets emailed twice by two different sequences. Deliverability degrades and nobody notices for a fortnight because the tool that sends is not the tool that monitors.

One system holding all of it can, in principle, close those gaps — it knows what it sent, to whom, from which domain, and how that domain is performing.

The honest caveat is that consolidation always trades depth for coherence. A specialist enrichment platform will do enrichment better than a bundled one; a dedicated deliverability service will monitor more carefully. You are buying fewer seams at the cost of some ceiling on each component, which is the same bargain every all-in-one suite offers.

Outbound rarely fails because the copy was bad. It fails in the joins between six tools, which is exactly what nobody is watching.

Deliverability Is The Constraint Nobody Budgets For

Any tool that makes sending easier runs into the same wall, and it is worth understanding before you evaluate any product in this category.

Email providers have spent two decades getting good at detecting bulk unsolicited mail. Volume from a new domain, low reply rates, spam complaints, and identical message bodies are all signals that route mail to a folder nobody opens.

The cruel arithmetic is that a tool which lets one person send ten times more email does not produce ten times more meetings. It produces more sending, worse engagement metrics, and eventually a domain that cannot reach an inbox at all — including for the humans who share it.

This is why deliverability appears in Artisan's own list of consolidated functions, and why it is the right thing to interrogate hardest in a demo. Ask specifically: does it use separate sending domains, how is warm-up handled, what happens to sending when reply rates fall, and can it be configured to stop rather than push through.

The signal to look for is a product that limits sending on your behalf. A system with no brakes is not more powerful — it is a faster route to a burned domain.

Personalisation helps here and only partially. Genuinely researched outreach earns replies; personalisation tokens filled from a database do not, and recipients tell the difference immediately.

1. Run it on a dedicated sending domain, never your primary corporate domain.
2. Start at low daily volume and hold there through a full warm-up period.
3. Read the first fifty messages it writes before any of them send.
4. Watch reply rate rather than open rate — opens are unreliable and easy to flatter.
5. Agree in advance what reply rate triggers a pause, and confirm the platform can enforce it.

Reading The AI Employee Framing

Artisan names its products Artisans and gives them human names. That is a deliberate positioning choice and it is worth being clear-eyed about what it does.

The framing is useful in one respect: it sets the right expectation about scope. Ava is not a feature you add to a workflow, it is an attempt to own an entire function end to end. Buyers who evaluate it as a sequencer will misjudge it.

The framing is misleading in another. An employee exercises judgement, learns your market by talking to people in it, notices that a segment has gone quiet and asks why. An agent executes a defined loop very consistently. Those are different things, and the gap shows up specifically in strategy — deciding who to target and what to say — rather than in execution.

In practice the good outcome is not that you stop needing a person. It is that one person can own the strategy for what previously took a team to execute: choosing segments, writing the positioning, reading what came back, and adjusting.

That is a genuinely valuable change and it is a smaller claim than the marketing makes. Teams that buy it expecting to remove headcount rather than redirect it tend to be disappointed.

Assisted rather than autonomous tools like Amplemarket make a smaller version of the same bet. The backing suggests the market takes the category seriously — a $25 million Series A led by Glade Brook Capital, with HubSpot Ventures participating, which is a notable investor for a product living next to a CRM.

Credits, Not Seats

This site does not publish plan prices. The structure matters more anyway, and Artisan's is genuinely different from the tools it replaces.

Artisan charges on a credit-based model, with tiers spanning startup through enterprise. You are buying the agent's capacity to work rather than a licence for a person to log in.

That is the correct shape for the product and it changes how you budget. Seat-based tools cost the same whether a rep sends ten emails or a thousand; a credit model scales with activity, which means cost tracks volume rather than headcount.

It also means the failure mode is different. With seats, waste looks like an unused licence. With credits, waste looks like capacity spent on a badly chosen segment — you can burn a month's allocation on outreach to the wrong people far faster than a human rep could.

Three things to establish before signing: what specifically consumes a credit (contacts sourced, emails sent, enrichments, replies handled — they are not equivalent), whether unused credits roll over, and what the annual commitment is. Enterprise tiers in this category are usually annual.

Because targeting quality now directly determines cost efficiency, the discipline that used to be about protecting a rep's time is now about protecting your balance.

Where A Human Rep Or A Specialist Still Wins

Being specific here is more useful than another feature list.

Complex, high-value, few-target sales. If your total addressable market is 200 companies and each deal is transformative, automated outreach is the wrong instrument. Those relationships are built by a person who knows the industry, and volume tooling actively damages them.

Anything requiring genuine domain credibility. Selling to clinicians, engineers or lawyers means the first message has to demonstrate that you understand their work. Generated research reads as generated to a specialist audience faster than to anyone else.

When your problem is targeting, not sending. If you do not yet know who buys from you, automating outreach scales a guess. Clay and similar data-first tools are built for working that out.

Regulated and consent-based markets. Cold outreach obligations differ sharply by jurisdiction — the rules covering unsolicited commercial email in the EU and UK are materially stricter than in the US, and automation does not change who is responsible for compliance.

• Enterprise sales with a small named-account list
• Technical audiences who detect generic research instantly
• Companies still working out their ideal customer profile
• Markets where cold email consent rules are strict
• Any motion where the relationship, not the volume, is the asset