AI Transformation · babaVC
We design an AI team around how your business works, build it into your tools and lead it until it runs. Bring David in as your Fractional Chief AI Officer, or run it as a fixed-scope Transformation project.
1Requests come in
2Agents work
3Checked, approved
Queue clear
Nothing waiting
4Shipped, watched
Invoice copy, ticket #4817
Why now
Today each person prompts on their own and starts from zero, so what one of them learns stays with them. On a platform the work, the context and the skills are shared, so each run can build on the last. In McKinsey's State of AI 2026, 80% say AI improved their productivity, but only 37% of organisations report any EBIT contribution.
Tickets, reviews, drafts and checklist runs leave your builders' desks.
One view of agent spend, under a cap you set.
The work keeps moving overnight. Anything you gate waits for your people in the morning.
Nothing you gate ships without a person.
An AI team takes on the repeat work in five lanes, from the first support reply to the weekly report.
Slack · #support
Customer can't log in after a password reset. Same thing as yesterday.
Support
Drafted a reply with the reset steps for their plan and a link to the help page. This matches yesterday's ticket, so I opened a bug for Builder with the steps to reproduce. The reply is waiting for your OK.
Waiting for review · Support lead
85% of customer service leaders are expanding human agents' responsibilities as AI takes on contact volume; just 31% are planning frontline layoffs.
Gartner, Apr 202687% of customers say a company using GenAI for service must still give them access to a human.
Gartner, Aug 2026Schedule · Mon 08:00
Follow up on every lead that went quiet after their demo.
Growth
Found the leads with no reply since the demo and drafted a short follow-up for each, built on what they asked on the call. The ones that mention a competitor are flagged for you. Outbound email is on your approval list, so nothing is sent yet.
Drafts ready · Account manager sends
54% of sellers have used agents, and nearly 9 in 10 plan to by 2027. They expect agents to cut prospect research by 34% and email drafting by 36%.
Salesforce State of Sales, Feb 2026By 2028 AI agents will outnumber sellers 10 to 1, yet fewer than 40% of sellers will say agents improved productivity; CSOs who fix data, automation and UX first are 5× more likely to see ROI.
Gartner, Jul 2026GitHub · PR #88
Ready for review: new export endpoint.
QA
Checklist run, tests pass. One problem: with an empty filter the endpoint returns other users' email addresses. Sent back to Builder with the line and a fix. Not merged.
Sent back · Builder
Organisations that use AI and automation extensively in security saved $1.93M per breach compared with those using none.
IBM Cost of a Data Breach 2026AI agents solved cybersecurity problems 93% of the time, up from 15% in 2024.
Stanford HAI AI Index, Apr 2026Linear · spike
Can customers book a demo straight from the pricing page?
Builder
Built it on a branch: a button on the pricing page, a booking form and the booking email as a draft. Preview link is in the issue. Changes reach production when you merge them.
Preview ready · Product lead
On SWE-bench Verified, a key coding benchmark, performance rose from 60% to near 100% in a single year.
Stanford HAI AI Index, Apr 202632% of organisations skipped buying at least one software product because they built it in-house with agentic coding tools.
McKinsey State of AI, Aug 2026Schedule · Mon 08:00
Weekly report for the leads channel.
Ops
Pulled last week's closed tickets and open bugs from Jira and the merged PRs from GitHub. Report drafted in Notion. It posts to #leads after your OK.
Waiting for review · Ops manager
Active agents in Microsoft 365 grew 15× year over year, and 58% of AI users produce work they couldn't a year ago.
Microsoft Work Trend Index, May 2026Customer service leaders' AI spending rose 38% while their overall budgets grew 2%. Gartner's line on why: “Organizations are moving beyond isolated AI use cases and toward connected ecosystems of AI agents.”
Gartner, Aug 2026Either way it starts the same: context first, agents on real work from week two. After that, David keeps driving it as your Fractional Chief AI Officer, or your team raises tickets and we support it.
Weeks 1–2
We map how work gets done across your tools and write it down, so the agents know your company before they start.
From week 2
The first agents work on real tasks from week two, and we test, evaluate and tune them weekly against your work until they're reliable.
Ongoing
As your Fractional Chief AI Officer, David keeps driving the team, the roadmap and the rollout across the company. On a fixed-scope project, your team raises a ticket when something needs to change and we fix, tune or extend.
Platform
Paperclip runs agents the way a company runs people: each one has a role in an org chart and a budget of its own, your company configures the approval gates, and runs are written to an audit log. It works with the agents you already use, Claude Code, Codex, Cursor and more, and it's open source under the MIT licence.
We set it up in your own environment, connect your stack, build the team and support it.
GitHub and npm as of . Discord as of .
“Those 75,000 employees will be working with 7.5 million agents.”
“If those systems are fragmented, the agents will scale the fragmentation.”
“The real advantage comes from combining AI efficiency with human judgment, empathy and experience.”
Humans approve whatever you choose to gate: outbound email, merges, publishing. Change the list whenever you want.
Model usage is billed by the provider straight to your account. Spend caps and stop switches live in the org.
Slack, Teams, GitHub, Linear, Jira, Google Workspace, Notion and more, plus the harness your builders already use.
Your data stays in your own environment, and the agents run on AI provider keys you hold.
Fifteen years in tech, building products and scaling organizations. Jack of all trades: tech, product, business.
Co-founded EasySize (AI for fashion sizing, from scratch to PMF and profitability over eight years). Most recently ran tech, product and strategy at Cludo (site search product, company-wide AI rollout, the best-selling quarter in the company's history).
Click or tap to flip
You work with David Babayan: fifteen years building products and scaling organisations, EasySize co-founder, most recently tech, product and strategy at Cludo.
Anything you gate waits for a human, who can approve it or send it back with a note. The agent redoes the work from that note. If the same mistake keeps coming back, raise a ticket and we tune the agent.
Slack, Teams, GitHub, Linear, Jira, Google Workspace, Notion and more. If a tool has an API or an MCP server, it's likely in. If you use something unusual, we'll check it on the call.
Just Paperclip, and we set it up for you in your own environment. Your people stay in the tools they already use, and the agents work there too.
You do, per agent and per action. During the build we recommend gates, such as outbound email, merges and publishing, and you set them. You can add or remove a gate at any time.
The first agents work on real tasks from week two, and the build-and-iterate phase runs to about week six. You see their output on your own work while the build goes on.
A Transformation project gets a fixed-scope quote at the end of the context sprint, once we know which work the team will take on. A Fractional Chief AI Officer engagement is a monthly retainer, agreed on the call. Model usage is billed straight to your own provider account. You set a spend cap per agent, and work pauses at the cap until someone raises it.
Keep them. Those are assistants one person drives, one chat at a time. The AI team adds shared context, roles and gates on top, so work is picked up from your tools, handed between agents and checked. Your builders keep Cursor, Claude Code, Codex and more inside it.
No standing access unless you want it. The team runs in your own environment on your keys, so it doesn't depend on us. Support runs on tickets you raise, and you decide what access we get for each one.
We set it up in your own environment, with your access controls. Runs are logged and the code is open source, so your team can read it.
Yes. The team lives in your own environment, so you keep the agents and their setup whether or not you raise another ticket with us.