W’adwuma Foforo Ne Sɛ Wobɛhyɛ AI Agents mu: Sɛnea AI Native Companies Yɛ Adwuma Ankasa

In This Article

Dear subscribers,I’ve spent the past few months bisabisa akannifo nsɛm wɔ AI-kurom nnwumakuw mu. I’m now convinced that:Onboarding and managing AI agents IS the job, no matter what your function is.Wɔ saa free deep dive yi mu no, mepɛ sɛ mekyɛ sɛnea nnwumakuw abiɛsa a wɔyɛ AI-kuromfo —Linear, Ramp, ne Factory — de nnyinasosɛm yi di dwuma. Nsɛm bi a wɔafa aka afi emu biara mu ni:Nan Yu (Head of Product, Linear): “Wobɛnya AI kuw no mufoɔ a wobɛtumi de nnwuma ahyɛ wɔn nsa na wɔne wɔn akasa te sɛ sɛdeɛ wo ne nkurɔfoɔ kasa no ara pɛ.”Geoff Charles (CPO, Ramp): “Sɛ womfa Claude Code nni dwuma a, ɛmfa ho nea wo dwumadie yɛ a, ebia wonyɛ adwuma yie.”Eno Reyes (CTO, Factory): “We codified product management, frontend UI, data nhwehwɛmu, ne nea ɛkeka ho kɔ ahokokwaw a wotumi de di dwuma bio a odwumayɛni biara betumi afrɛ no mu.”Kenkan kɔ so ma wohwɛ sɛnea AI-kurom adwumakuw biara yɛ adwuma wɔ mu.Linear: Making AI agents first-class teammatesYou can @mention an agent to create and assign an issue to another agentLinear’s approach to agents is shaped by their product. Nan (Linear’s Head of Product) gye di sɛ:Ɛsɛ sɛ Agents yɛ adwumayɛfo a wodi kan. Ɛsɛ sɛ wutumi de ka nnwuma ho, de wɔn hyɛ nsɛmpɔw mu, na woka ho asɛm wɔ nsɛm a wɔka mu.Nanso, Nan gye di nso sɛ onipa no kɔ so bu nea ebefi mu aba no ho akontaa bere nyinaa. Here’s how Linear builds products with agents wɔ anammɔn biara mu:Te ɔhaw no ase. Agents kenkan na wɔbɔ adetɔfo nkɔmmɔbɔ biara a efi Intercom, Zendesk, ne Gong mu no mua. Wɔn auto-create nsɛm, de-dupe wɔn tia backlog, na wɔde ma kuw a ɛfata. Kyerɛ ano aduru no. Esiane sɛ agents wɔ kwan kɔ adetɔfo nkɔmmɔbɔ so nti, wobetumi aboa wo ma woayɛ iterate wɔ spec bi so denam data-backed insights a wobɛtwe afi akwan pii so. Yɛ nhyehyɛe bi. Agents betumi abubu wo spec no mu ayɛ no concrete tickets na wɔde wɔn akɔ teams a ɛfata no so automatically. Wɔ Linear no, mprempren ananmusifo na wɔyɛ tekiti dodow no ara. Onipa adwuma ne sɛ ɔbɛsan ahwɛ wɔn adwuma mu na wayɛ nsakrae wɔ nsɛm a ɛfa ho no mu bere a bere kɔ so no.Execute. Mfomso ne nneɛma nketenkete nya de ma agents te sɛ Codex ne Cursor tẽẽ. Wɔ nneɛma a ɛyɛ den ho no, mfiridwumayɛfo fi ase Claude Code na wɔde Linear MCP di dwuma de twetwe nsɛmma nhoma no mu nsɛm a ɛfa ho nyinaa.Efi Nan: Ɛte sɛ nea nea agents betumi adi ho dwuma no kɛse retrɛw wɔ afe nkyem abiɛsa biara mu. Models foforo ne harnesses repia ɔhye no afi fixes a ɛnyɛ den so akɔ projects a ɛreyɛ den kɛse so.Want to build with agents like Linear? Ɛha na anammɔn 4 a mfaso wɔ so a Nan kyɛe wɔ nea wo kuw no betumi ayɛ nnɛ:Ɛsɛ sɛ developer biara default kɔ agentic coding adwinnade a edi kan. Eyi ne anammɔn a edi kan a ɛyɛ mmerɛw sen biara. Fa aban adwinnade (Cursor, Claude Code, anaa Codex) ma na hwɛ so sɛnea ɛbɛyɛ a wubetumi ahu utilization.Supplement with an async cloud coding agent. Async background agents tumi one-shot nsakraeɛ nketewa dodoɔ no ara ne bug fixes. Cursor ne Devin wɔ afɔrebɔ pa wɔ ha.Si so dua sɛ designers ne PMs nyɛ adwuma tẽẽ wɔ codebase no so. Agents te sɛ Claude bue kwan a ɛnyɛ friction ma PM ne designers sɛ wɔbɛyɛ nsakrae tẽẽ wɔ codebase no mu. Ɛsɛ sɛ obiara bɔ mmɔden sɛ ɔbɛyɛ ɔdansifo.Ɛsɛ sɛ PM ne aguadifo default to AI interface. Ɛsɛ sɛ saa dwumadie yi yɛ wɔn adwuma 80-100% denam chat interface so — sɛ ɛyɛ Claude, ChatGPT, Notion AI, anaa biribi a ɛte saa. Nan hu daakye bi a nnipa ne agents bɛyɛ adwuma abom wɔ spec level — akyerɛkyerɛ nea ɛsɛ sɛ wɔkyekye ne nea enti a — na afei wɔde nneɛma akɔma agents ma wɔadi biribiara a ɛwɔ downstream.Ramp: Measuring AI proficiency in 4 levelsRamp is rapidly pushing their employees to be AI-native in 4 levelsSɛ Linear kyerɛ sɛnea wobɛma agents ayɛ first-class agents part of your team, afei Ramp kyerɛ sɛnea wobɛma wo adwumakuw no nyinaa agye wɔn.Wɔ afe 2025 mu no, Ramp de nneɛma 500 + kɔmaa, oduu $1B sika a wonyae, na ɔyɛɛ ne nyinaa ne 25 PMs.Wɔyɛɛ eyi denam hwehwɛ a wɔhwehwɛe sɛ adwuma biara (eng, product, design, sales, marketing, legal, finance) bɛkɔ onboard na wɔne agents.Geoff (CPO Ramp) kyɛ nhyehyɛe bi a wɔde bɛsɔ AA ahokokwaw ahwɛ ama obiara odwumayɛni a mihu sɛ mfaso wɔ so kɛse:L0: Ɛtɔ mmere bi a ɔde ChatGPT di dwuma. Ɛda adi kɛse sɛ saa nkurɔfo yi rentra adwumakuw no mu bere tenten. Sɛ wonyɛ obi a ɔhyɛ wo ho ase a wowɔ onyin adwene wɔ AI nnwinnade ho a, Geoff se ɛbɛyɛ den yiye sɛ wobɛtete wo ma woadi mu.L1: Ɔde GPT, nnwuma, ne AI nnwinnade a ɛwɔ mu di dwuma na ɔsesa. Saa nkurɔfo yi resɔ AI ahwɛ nanso wonnya nyɛɛ adwuma ankasa automated.L2: Builds an app that automates part of theiradwuma. Saa nnipa yi betumi de koodu ahyɛ bɔ anaasɛ wɔde nsɛm a ntease wom ama wɔ nnipa afoforo adwuma ho denam AI nnwinnade so.L3: Systems builders. Saa nnipa yi rekyekye AI infrastructure ne ahokokwaw a ɛma obiara a ɔka kuw no mu yɛ ntɛmntɛm.The company’s goal is to push everyone up the ladder. L0s ankasa paw fi mu. L1s bɛyɛ L2s. L2s bɛyɛ L3s. Na L3s nya ahyehyɛde no mufo a aka no so nkɛntɛnso. Geoff nso kyɛɛ anammɔn 5 a adwumakuw biara betumi atu de abɛyɛ AI-kuromni:Yi friction nyinaa fi hɔ. Ma kwan ma AI nnwinnade a agye din a anohyeto biara nni token anaa sikasɛm nhyehyɛe ho na yɛ AI ahokokwaw a obiara betumi atwe afi mu akorae wɔ mu. Sɛ setup no yɛ den a, nnipa dodow no ara rennye adopt.Make adoption visible. Yɛ ɔmanfoɔ Slack akwan a nkurɔfoɔ bɛtumi akyɛ deɛ wɔasi. Wɔ nsa nyinaa mu no, kyerɛ wɔn a wɔnyɛ mfiridwumayɛfo a wɔreyɛ nneɛma a ɛyɛ nwonwa, te sɛ sikasɛm a wɔde si sikakorabea nhyehyɛe anaasɛ aguadi a ɛma wɛbsaet adebɔ yɛ adwuma wɔ ɔkwan a ɛyɛ adwuma so.Fa nsaano mmoa ma. Host office hours a obiara betumi de ne ho ahyɛ mu de akyekye AI nimdeɛ ne adwumayɛ nhyehyɛe. Wɔapaw AI ho abenfo a wɔn adwuma nyinaa ne sɛ wɔbɛka asɛmpa, ama nkurɔfo asiesie wɔn ho, na wɔaboa wɔn ma wɔadu “aha bere” no ho. Track sɛnea wɔde di dwuma no na fa wo ho gye mu. Ramp di token a wɔde di dwuma wɔ AI nnwinnade nyinaa so wɔ odwumayɛni biara mu. Akannifoɔ kyɛ saa data yi de yɛ abɔdeɛ mu akontabuo na wɔde wɔn ho hyɛ mu berɛ a obi dwumadie sua.Ma ɛnyɛ adwuma mu ahwehwɛdeɛ. Mprempren PM nsɛmbisa no ka nhyiamu a wɔatu ho ama a ɛhia sɛ woyɛ adeɛ a ɛyɛ adwuma na afei wokyerɛkyerɛ nea enti a wosii ne sɛdeɛ ɛyɛ adwuma mu. Geoff bɔɔ n’akannifoɔ nyansapɛ mua maa dwumadie biara wɔ Ramp wɔ line baako mu:W’adwuma ne sɛ wobɛyɛ w’adwuma no automate.From Geoff: “Sɛ meka kyerɛ me kuw no mpɛn 10 sɛ ɛsɛ sɛ CTA no wɔ soro fold no a, fix no nkyerɛ sɛ ɛyɛ 11th time. It’s encoding that feedback into an automated design crit process or AI skill so that it never happens again.”Factory: AI kuromni fi da a edi kanSɛ Linear ne Ramp kyerɛ sɛnea nnwumakuw gye AI ananmusifo a, Factory kyerɛ nea ɛba bere a wokyekye twa wɔn ho hyia fi da a edi kan no. Factory yɛ nnipa 55 AI software nkɔso adwumakuw a ne bo yɛ $300M a wɔahyehyɛ no atwa AI ho ahyia fi mfiase. Nea ɛma wɔyɛ soronko ni:Hye product engineersFactory nfa PM ne engineers wɔ ɔkwan soronko so. Mmom no, wɔfa nneɛma ho mfiridwumayɛfo a wɔhwɛ AI adwumayɛfo so na wɔne wɔn yɛ adwuma. Da a wɔtaa yɛ no te sɛ:Hwehwɛ traces a efi agent runs mu hwɛ baabi a nhyehyɛe no sii gyinae a enye. Kyerɛw fixes ɛnyɛ sɛ code, na mmom sɛ nniso (e.g., update to a skill, a new lint rule, or a refined automation)Hwɛ PR ahorow a agents frankaa sɛ asiane kɛse wom nkutoo mu (agents hwɛ nea aka no so).Suggest new ideas and debate prioritization with colleagues and AI.This work isn’t easy and requires deeper expertise, but the leverage is enormous.Make your codebase agent-readyAgents hia codebase a wobetumi ayɛ adwuma wɔ mu ankasa na ama wɔatu mpɔn. Factory scores codebases wɔ maturity levels anum mu, na Level 3 ("Standardized") ne baabi a ɛsɛ sɛ akuw dodow no ara di kan de wɔn ani si so.Factory’s agent readiness frameworkEncode expertise into reusable skillsSɛ wo codebase no yɛ agent-ready a, anammɔn a edi hɔ ne sɛ wobɛma agents nimdeɛ a wɔde besi gyinae pa denam ahokokwaw so (titiriw no, text markdown fael ahorow nkutoo). Factory de ahokokwaw di dwuma de kyerɛw animdefo ne adwumakuw nimdeɛ kɔ biribi a ɔnanmusifo anaa odwumayɛni biara betumi de adi dwuma mu. Here’s a list of skills that Factory uses internal and links to their markdown files ma wo sɛ wobɛkyerɛw na woasesa:Product management skill. Aguadeɛ nnyinasosɛm, nsoromma 11 osuahu nhyehyɛeɛ (wɔfɛm fii Airbnb’s Brian Chesky hɔ), PRD nsusuiɛ, nkontabuo rubric, ne kasa akwankyerɛ nyinaa wɔ markdown fael baako mu.Frontend UI integration skill. Ɔkyerɛ Droid sɛnea wɔde Factory’s design system, routing conventions, ne testing standards bɛkyekye features.AI data analyst skill. Yɛ nhwehwɛmu nhwehwɛmu, yɛ mfonini ahorow, na yɛ akontaabu amanneɛbɔ denam Python abɔde a nkwa wom nhyehyɛe a edi mũ no so.Internal tools skill. Yɛ admin panels, support consoles, ne adwumayɛ dashboards a ɛwɔ access controls a ɛfata ne audit logging baked in.Vibe coding skill. Rapidly prototype new web apps from scratch with modern frameworks.Sɛ wubetumi de nea wo nnipa a wɔyɛ papa no nim no ahyɛ ahokokwaw mu a, enhia sɛ wobɛfa animdefo ama adwuma biara.Anamɔn 6 a wode bɛfa eyinom nyinaa ayɛ adwumaTo recap:Onboarding ne managing agents rebɛyɛ adwuma titiriw ama adwuma biara.Nneɛma asia a wubetumi de adi dwuma mprempren ni:EfiLinear:Default developer biara ma agentic coding adwinnade te sɛ Cursor, Claude Code, anaa Codex. Ma PMs ne designers nkɔ codebase no mu. Ma wɔmfa PRs ne ship code a wɔde agents di dwuma no nkɔma. Gyae routing nsakrae ketewa biara denam engineer.From Ramp:Susu AI ahokokwaw wɔ wo kuw no nyinaa mu. Ramp’s 4 levels framework ma wo nsɛmfua a wɔkyɛ ma baabi a nkurɔfo wɔ ne baabi a ɛsɛ sɛ wɔkɔ. Di AI dwumadie akyi na ma ɛnyɛ adwumayɛ akwanhwɛ fã. Worentumi mma nea wonsusuw no ntu mpɔn, na nneɛma a ɛkanyan obi ho hia. Efi Factory:Score wo codebase ma agent ahoboa. Fa Factory’s agent readiness levels di dwuma na te ase sɛ wo codebase no ayɛ krado anaa.Codify wo team’s expertise into reusable skills. Fa nea wo nnipa a wɔyɛ papa nim no hyɛ ahokokwaw fael mu na ma ɛnyɛ mmerɛw mma nnipa ne ananmusifo nyinaa sɛ wɔde bedi dwuma.Nea ɛsen ne nyinaa no, onboard agents te sɛ you’d onboard a human. Ma wɔn nsɛm a ɛfa ho, fa wɔn bata wo operational stack no ho, na ma onipa bi mmu akontaa wɔ wɔn outcomes ho.Ma menhu nea wususuw wɔ comments no mu!

You May Also Like

13 Trending Songs on TikTok in March 2026 (+ How to Use Them)

Our Best AI PM Workflows for Prototyping, Strategy, and Personal OS (2026)

China’s Cybersecurity Agency Issues Security Warning About OpenClaw