01 / Map
Where AI actually helps your business
We go function by function. In each we mark what pays off for a small team and how ready it is to use today.
Find, reach, prepare
- Account enrichment: who decides, reasons to reach outready
- Signal monitoring: incidents and breaches at similar companiesready
- Personalizing the first touch in bulksetup
- Meeting prep from past calls and CRMready
Volume and repurposing
- SEO articles and case studiesready
- One case study into ten formats: post, newsletter, landing blockready
- Drafts for newsletters and socialready
First line and triage
- Ticket triage and routingsetup
- Answers from the product knowledge basesetup
- Summarizing long technical threadsready
Productivity, not expertise
- Coding assist, docs, testsready
- Internal tools and proceduresready
- Your core product IP is your expertise - AI stays out of ityour zone
Contracts and summaries
- First-pass contract review (human finalizes)setup
- Reporting and summariesready
- Onboarding and internal processesready
Next: three build-outs that close these cells in practice. They can be shown live.
02 / Build-out
AI SDR - an outbound sales pipeline
Not a mail blaster, but the chain of who, on what signal, with what message. Exactly the layer that strengthens your funnel at find-and-reach.
From a list of companies to a meeting-ready lead
The human stays on decisions and the conversation. The machine takes collection, enrichment and drafts.
03 / Build-out
Content factory / article engine
A stream of SEO articles and materials without hand-writing each one. Covers the marketing cell: search visibility and repurposing expertise.
Keywords - drafts - review - publish
One expert on review instead of a copywriter team. Your expertise fuels the stream.
04 / Build-out
Turnkey AI agents and a live stand
So AI does not stay a slide: agents assembled for the task, and a stand where the chains run live. This is where you see it, not take it on faith.
Agents assembled for your process, shown on a stand
Ready-made skill bricks assemble into an agent for a routine: outbound, support, contract review. On the stand the agent answers from your data and shows where the answer came from - it does not make things up.
05 / Roles
What to hand to AI, what to keep with the human
AI drafts - the human decides and owns it
Hand to AI - draft, volume, routine
- First drafts of emails, content, documents
- Collecting and enriching account data
- Monitoring, summarizing, triaging tickets
- Generating options: ten angles instead of one
- First pass over code and contracts
Keep with the human - judgment, face, risk
- Final text that goes out under the company name
- Deciding who to call and when
- Client relationships, negotiation, price
- Choosing the option and owning it
- Security-critical and final legal calls
In 2026 you do not hand over a whole role. You hand over 60-80% of the routine inside a task. An AI employee replacing a department is marketing, not reality.
06 / Skills
What the team should know and where to start
So a small team does not scatter. Order matters: one person and one task first, then scale.
A practical first step: pick one painful routine, put AI on it with a human in the loop, measure the time. Not 'adopt AI in general'.
07 / Data & security
What may and may not go to the cloud
Separately, because some data is sensitive. Heavy reasoning lives in cloud models - so you need a rule for what goes there.
A sensitivity framework
Three data tiers - three modes
Public
Marketing, articles, open materials. Freely to cloud models.
Internal
Processes, drafts, non-personal data. To the cloud with a policy and anonymization.
Sensitive
Client and personal data. Local only or under tight control, not to shared cloud services.
In practice: reasoning goes to frontier models in the cloud, and only a light layer (OCR, small models) runs locally on modest hardware. An AI-use policy decides what goes where.