GTM by Blash
GTM by Blash is an AI outbound system for firms that sell to a narrow, specific kind of buyer. It qualifies the uploaded list against your ideal customer profile first, and drafts only for the contacts that fit.
Watch it work
13-step
path from onboarding through to deliverability
ICP-first
classification before a message is drafted
3 buckets
qualified, uncertain and drop, kept separate
Per prospect
a current signal and a chosen service, then the draft
A/B tested
variants set up before the sequence is built, not after
DMARC
warmup, auto-suppression and a domain reputation tile
01
Most outbound tools optimise the wrong end
Most outbound tools are built around send volume. More mailboxes, faster sequencing, better templates. The arithmetic looks sound until you apply it to a list that was never any good, at which point volume makes the problem larger and takes the sending domain with it.
GTM by Blash starts at the other end. Before a single message is drafted, it classifies the uploaded list against your ideal customer profile and removes the contacts that do not fit. In the read-only tour, which runs on illustrative data for a fictional B2B SaaS called Flowbase, 11,000 classified contacts resolve into 3,850 qualified, 2,200 uncertain and 4,950 dropped. Forty-five per cent of that demo list never receives anything at all.
Those are illustrative figures, not a client result. The point is the shape of the decision. A product that writes to fewer people has to be right about which people, and it has to show the reasoning that put each name in each group.
02
A thirteen-step path, and you can see all of it
The product is laid out as a numbered path in the left rail. It is deliberately linear, and each step is a point at which a contact either progresses or is removed.
Nothing is hidden behind a single button marked go. Each step is its own screen, so you can see what the qualification, the coherence review and the research step each did before anything was sent.
Steps 1 to 3. Onboard, find leads, upload and qualify. The ICP is defined and the raw list arrives.
Steps 4 to 5. ICP classification, then a coherence review before anything is drafted.
Step 6. Research and personalise, one prospect at a time.
Steps 7 to 9. A/B testing, sequence build, then warmup and cadence.
Steps 10 to 12. Sends and tracking, replies, then meetings and results.
Step 13. Deliverability, treated as its own discipline rather than a setting.
03
What the AI finds, one prospect at a time
Step six is where the AI does its most visible work. The screen carries the heading "What the AI found" against every prospect it kept, and the product explains itself in its own words: for each kept prospect the AI pulls a specific, current signal, picks the single best-fit service, and writes from that.
Three jobs, in order. Find one specific, current signal about this company. Pick the service that finding points at. Then draft.
Firms that sell several distinct services often lead with whichever one the writer is closest to that week. Selecting the service from the signal, per prospect, is a different discipline from swapping a first name into a template.
What comes out is a draft with the finding it was written from sitting beside it. You can read the finding and judge the draft against it.
04
Testing before sequencing, not after it
A/B testing sits at step seven, ahead of the sequence build at step eight. That ordering matters. The variants are set up before the cadence is built rather than argued over once the sequence is already running.
Warmup and cadence follow at step nine. New domains and new mailboxes ramp before they carry real volume, and cadence is a step in the path rather than a manual setting.
05
Deliverability is a first-class concern
Step thirteen covers DMARC, warmup, automatic suppression and domain reputation. The dashboard keeps reputation on the front page as a standing tile, next to the bounce rate, rather than burying it in a settings screen where nobody looks until the replies stop.
In the demo the bounce rate reads 1.1 per cent with 39 addresses auto-suppressed, and domain health reads HIGH. Illustrative figures again, but they show what the product puts in front of you. Bounced addresses are suppressed by the system rather than cleared by hand.
A sending domain is quick to burn and slow to rebuild. Any tool that treats deliverability as configuration rather than as a live number is asking you to find out the hard way.
06
Replies through to meetings
Steps eleven and twelve carry replies, then meetings and results. The dashboard carries a tile for each stage: contacts uploaded, qualified, in sequence, opened, clicked, replied, meetings booked, bounced, domain reputation, and a pipeline value.
In the demo the tiles read 3,740 contacts in sequence, 120 replies, 38 interested and 22 meetings booked. Those are fixture values on a synthetic list, not a conversion rate you should expect. The pipeline tile carries the word illustrative on its own face.
We are showing you the reporting shape, not a benchmark. Every stage is counted and every drop between stages is visible, which is what you need in order to see where the funnel is leaking.
07
Look at it before you speak to anyone
The tour at gtm.blash.uk is read-only. Its banner reads: read-only tour with illustrative data, viewing Flowbase, a fictional B2B SaaS. There is a four-minute walkthrough and the full thirteen-step path is clickable.
If it fits, we configure it around your ICP, the services you actually sell, and the way you sell them. The product is the same one you clicked through.
Common questions
How does the AI actually work?
You define the ICP. The AI sorts the uploaded list against it into qualified, uncertain and drop, and only the qualified group reaches the writing step. For each of those it finds one specific, current signal about the company or the person, picks the single service that best fits what it found, and drafts from that. The finding sits next to the draft, so you can see what the draft was written from. It classifies, researches and writes. It does not set your ICP and it does not decide commercial strategy.
What happens to the contacts it is not sure about?
They are held in their own group rather than pushed into the send. In the demo, 2,200 of 11,000 sit in the uncertain bucket, separate from the 3,850 qualified and the 4,950 dropped. A tool that resolved every ambiguity in favour of sending would be simpler to build and would put the sending domain at risk.
What does it not do?
It is not a CRM and does not try to replace one. It does not make calls, run the meeting or close the deal. It will not promise inbox placement, because nobody can promise that. And it does not present a figure it has not counted: the pipeline number in the demo is labelled illustrative on the tile itself rather than dressed up as a forecast.
How do you protect the sending domain?
Deliverability is its own step, not a checkbox. DMARC, warmup, automatic suppression and domain reputation all sit inside that step. New mailboxes and domains go through warmup before they carry cadence, bounced addresses are suppressed automatically, and domain reputation sits on the dashboard as a live tile. In the demo that reads 1.1 per cent bounced with 39 auto-suppressed and reputation HIGH.
How is this different from the sending tools we already use?
Those tools begin at the sequence. This begins several steps earlier, at the question of who should be written to at all, and it answers that question by removing people. If your list is already tight and correct, the difference is smaller. If it is long and of uneven quality, the qualification step is where the difference sits.
Is there something we can look at first?
Yes. gtm.blash.uk runs a read-only tour on a fictional company, with a four-minute video and all thirteen steps open to click through. Nothing on it is live and every number on the screen is illustrative. Look at the structure and the reasoning rather than the figures.
Next step
See it on your own numbers
A short walkthrough with the people who built it.
