Why we built a back-office for engineers, not teams

There is an unspoken rule in how AI is sold to engineers right now, and once you see it you cannot un-see it. Every product in the adjacency — every one — has a VP on the other side of the contract. Trigify sells social-signal intelligence to GTM heads. Taplio sells content automation to marketing leaders with team-seat rollouts. Otter sells meeting transcripts to revenue ops. Shield sold LinkedIn analytics to the same buyer before LinkedIn cut its access and shut it down. The names rotate; the buyer never does. The engineer is not the customer. The engineer is the data source the buyer is paying to see.

We built SideKyk for the other side of that contract. Our buyer is the individual engineer — a salaried software engineer with two kids and a thirty-minute commute — the one who has been pitched fourteen AI productivity tools by their VP this quarter and has installed exactly zero of them. That choice — who pays — is not a marketing decision. It is the entire architecture of the product. The rest of this piece is about why.

The category nobody is building, and the one everyone is

Walk through any list of AI tools aimed at the tech professional and the pattern is loud. Trigify is "social signal intelligence for GTM teams." Taplio is "the LinkedIn growth platform" sold per seat to sales, marketing, and "team-seat" rollouts. Extrovert and Engage AI promise to help sellers warm prospects with AI-drafted comments. Even the ones with no engineer-as-buyer pretence — the Otters and the Notebook AIs and the various "meeting intelligence" products — get bought by a director who needs reporting up.

What every one of those products has in common is a buyer who is not the person doing the work. The VP buys; the engineer is enrolled. The unit of value is "team productivity" — a phrase that, decoded, means "what the VP can see about the engineer's output." This is why every one of those products eventually grows a dashboard, an analytics tab, a leaderboard, a Slack digest summarising whose pull-request velocity dropped this sprint. The dashboard isn't a feature. The dashboard is the buyer's actual purchase.

The category nobody is building is the inverse. A tool whose buyer is the engineer themselves — paid out of pocket, or out of an employer's learning budget the engineer controls — and whose only obligation is upward to that one person. No leaderboard. No VP-facing report. No quarterly business review where someone explains why Maya's "engagement score" is down. The tool serves Maya. The end.

We are not the first people to notice this gap. We are, as far as we can tell, the first people to deliberately design a product around it.

Who the buyer is, and why it changes everything

Maya is a senior DevOps engineer at a regulated fintech, eight years in, K8s and Terraform and a little Go, two kids aged six and nine, thirty-minute commute. Her quiet windows in a day are the commute, lunch if she can grab it, and forty-five minutes at nine PM after the bedtime routine. She has bought zero subscription tools for her career in the last three years. Not because she doesn't value her career. Because every tool that gets pitched to her either runs through her employer's procurement (in which case her VP gets the read receipts) or runs against her Slack and Jira (in which case her work knows what she searched for at 9:14 PM on a Sunday).

The B2C bet stated plainly: we sell to Maya. The pricing is salaried-engineer-affordable, not enterprise-seat-priced. The channel is WhatsApp, which is already where Maya lives — not another dashboard she has to remember to open. Cancellation is one message to her concierge: "cancel, please", and that's it. No retention call, no procurement reversal, no awkward conversation with her VP about why she stopped using the thing the company didn't buy.

This sounds like a pricing decision and a channel decision. It is actually a power decision. When the buyer is the user, the product's incentives compress to a single thing: was this hour of attention worth it to this one person this week? When the buyer is a VP, the product's incentives explode outward: collect more data, surface more reports, justify more seats, expand more surface area. The DAU model and the per-seat model both push the same direction — make the user spend more time inside your product, because that is the metric the buyer pays for. We push the opposite direction. We are paid to compress the engineer's attention to the three things that actually matter and then get out of the way.

A consumer subscription priced for a salaried engineer cannot afford to waste the engineer's time. A B2B contract priced per seat is structurally indifferent to whether any individual seat is well-served, as long as the org renews. The math is the moat.

The privacy boundary that comes with that buyer

Once you decide the engineer is the buyer, the rest of the architecture falls out fast. The most consequential decision — the one that took us the longest to write down because it felt counter-intuitive — is the one about what we are not allowed to read.

SideKyk has no access to Slack. No access to Jira. No access to corporate email. No access to the company VPN, the internal wiki, the build system, the dashboards, the on-call rotation. The data surface is strictly first-person and strictly consented: the engineer's public posts, their highlighted articles, the conference talks they tell us they're going to, the PR descriptions they choose to share. That is the whole list. We will not negotiate this.

Why does this matter? Because every B2B-DevEx tool pitched at engineers eventually leaks. The Slack integration installed for "team productivity insights" surfaces in a dashboard the VP didn't even know to ask for. The Jira read access added for "automated standup summaries" generates a report that shows up in someone's all-hands deck. We have watched this pattern repeat enough times that we treat the integration request itself as the warning sign. The minute the privacy boundary is even negotiable, the trust contract is already gone.

The B2B-DevEx tool cannot ship our model. Not because the engineers there are less thoughtful. Because the buyer — the VP who signed the contract — would not pay for a tool whose explicit promise is that it ignores their visibility needs. The B2B price tag is purchased with the engineer's data surface. We refuse the price tag so we can refuse the surface. This is the load-bearing thing.

When Maya pays us herself, she gets to set the rule: you can read what I show you and nothing else. When her company pays for the tool, the rule belongs to her company, no matter what the marketing page says.

Why "back-office" is the right metaphor, and "AI assistant" is not

There is one more decision we want to defend in public, because we keep getting asked about it.

The AI tools market has converged on a vocabulary — "AI assistant," "AI copilot," "AI agent" — that we deliberately don't use for what we ship. Every one of those words evokes a single function. An assistant schedules. A copilot completes code. An agent runs a task. They are the metaphors of seat replacement — a thing that does one job a person used to do.

That is not what Maya needs. Maya does not need one more thing that does one thing. She has Pluralsight for skill paths, ChatGPT for ad-hoc questions, Taplio (if her company bought it) for content, Otter for transcripts, a newsletter for news, Levels.fyi for comp data. The point-tool sprawl is the actual problem. Stacking one more point tool — even a clever AI one — makes her life worse, not better.

What Maya needs is what a VP already has. A VP has a Chief of Staff. The Chief of Staff is not a function; the Chief of Staff is an office — a small team that runs in the background, reading the firehose, prepping the deck, drafting the note, flagging the opportunity, keeping the calendar honest. The Chief of Staff's office is the integration. The role exists because the work doesn't decompose into single-function tools.

We are not building Maya an AI assistant. We are building Maya the office of a Chief of Staff, compressed into WhatsApp. Pulse is her intelligence analyst, reading the eleven hundred Hacker News stories so she reads the three that matter to her stack. Growth is her career strategist, watching the market signals around her trajectory and turning the nine PM "is DevOps dying" dread into a twelve-week bridge plan with a learning lesson that lands at 7:30 the next morning. Voice is her communications lead, drafting in her voice from the actual work she did this week — not generic LinkedIn slop she would be embarrassed to ship. Companion is her briefing aide, sitting beside her in every vendor demo and webinar and whispering the right question, then turning the talk she watched at 9 PM into two post drafts before bed.

No single B2B point tool gets to be all four. That is precisely the integration that makes the back-office worth more than the sum of its pieces. The relationship compounds. Month-three Maya is worth ten times month-one Maya, because the office remembers, and it gets measurably better at serving this one specific person.

The honest limit

We owe one honest sentence here, because the alternative is to oversell. SideKyk will not replace a three-hundred-dollar-an-hour human career coach. It will not give Maya what an actual Chief of Staff would give a CEO — the human judgment, the long lunch, the off-the-record warning that her director is about to leave. We are not making that claim.

What we are claiming is narrower and more honest. We are the office of a Chief of Staff that a salaried engineer with no spare evenings can actually afford to staff. The price point is the entire bet. A coach who charges what a coach should charge is not a tool a parent with two school fees and a CFP next month can run for themselves on a Tuesday night. We are. That is the whole product.

The wedge

So here is the wedge, said plainly. We are not selling your VP a way to see what you are doing. We are giving you the staff your VP gets — at a price you can afford, on a channel you already live on, with a privacy boundary that makes the trust contract structurally different from any tool your company will ever buy.

The aggregators in this space cannot ship this. Their incentive is to expand surface area so the buyer keeps renewing. Ours is to compress your attention to what matters this week. Trigify and Taplio and Shield and the next ten dashboards that come for the same buyer will keep getting funded, because their buyer keeps paying. We are betting that the engineer — the one with two kids and a thirty-minute commute and forty-five free minutes at nine PM — is the better customer to serve. Not because they are easier. Because the contract is honest, and the trust compounds.

So what now

SideKyk for tech professionals is in waitlist. If the back-office you have been wishing for sounds like the one we are describing — yours, in WhatsApp, no Slack and no Jira, the office instead of one more app — drop your number at sidekyk.ai/tech. We will text you when your spot opens. One message in, one message out. That is the relationship we are signing up for.