Mime-Version: 1.0 Content-Type: multipart/alternative; boundary="5ba85d8c17e17b21ec2f68599823247667e7978b459ab2535e98ccea49a3" Subject: Case Study: Moving from AI-First to AI-Native From: Carilu Dietrich from Hypergrowth Leadership To: Hidden Recipient Date: Tue, 30 Sep 2025 15:37:48 +0000 X-Hiring: We are hiring, reach out at header-hacker@emailshot.io X-EmailShot-Signature: KEf05_uwP1zdV3hTsWsof4JGbs2qnOkgdzueXubPvUlz1-xd_xzCpEj2_5ydU6Dj_r-uI-oDVHevaRBOBjssAw== --5ba85d8c17e17b21ec2f68599823247667e7978b459ab2535e98ccea49a3 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable View this post on the web at https://www.carilu.com/p/case-study-moving-fro= m-ai-first-to What is the difference between =E2=80=98AI-Forward=E2=80=99 or =E2=80=98AI = First=E2=80=99 and truly =E2=80=98AI Native=E2=80=99? Can you get native if= you weren=E2=80=99t born that way?=20 At the 10X CEO Conference two weeks back, Prukalpa =E2=9A=A1 [ https://subs= tack.com/redirect/163340de-92c5-45c4-8fb1-34c6ea2ccb26?j=3DeyJ1IjoiM2dmeXZt= In0.xu76uFObqArDfP822j-jnN48_jCfgM3m0rbAsF0l24U ] Sankar, founder and co-CE= O of Atlan, laid out a detailed roadmap and toolset of their journey to bec= oming AI-Native that was impressive. =20 After substantial focus and push, this company of 400 has reimagined roles = and jobs to be done, built more than 152 agents that made 4,000 runs in jus= t 5 weeks. They have massively sped up processes and unlocked new responsiv= eness that wasn=E2=80=99t possible before, especially in customer support. = The company is AI-ifying across every role and toolset. I share her tech st= ack, process, and takeaways below. For her:=20 'Being "AI First" means AI is considered as the primary solution. But being= "AI Native" means AI is fundamental=E2=80=94integrated into the fabric of = our thinking, operations, and processes. It means reimagining workflows fro= m first principles in a world powered by AI, rather than simply layering AI= onto existing processes.=E2=80=9D Four-Part AI-Native Roadmap Atlan took a four-part journey on its sprint to AI-Native: Phase 1: AI Task= force, Phase 2: Shift the culture, Phase 3: Embed into hiring, Phase 4: Ret= hink the org chart. Phase 1: AI taskforce=20 Prukalpa=E2=80=99s first step, one many teams have taken, was to assemble a= n elite team of people who had already shown passion and skill in AI. They = were tasked with taking the most impactful company use-cases live, determin= ing company-wide tooling, and driving enablement.=20 Phase 2: Shift the culture She then launched an AI Native mission to define overall strategy across th= ree long-term pillars:=20 Next came an AI productivity challenge to the company - a time-bound contes= t where team members could win prizes and be celebrated for identifying whe= re they could leapfrog their productivity with AI. More than 300 people con= tributed, and AI started entering workflows across the business.=20 Phase 3: Embed into hiring Next, AI became part of the hiring process. Atlan introduced hiring veto po= wer for candidates who didn=E2=80=99t have enough AI curiosity or AI skill.= In the interviews, Atlan made it explicit that candidates were expected to= use AI extensively, even and especially in interview project rounds. They = even asked candidates to open up AI and prompt in front of the interviewer = to see how they interact and think using AI. Their guidance to the team on = interviewing for AI skills:=20 Phase 4: Rethinking the org chart=20 Then, Atlan went on to rethink the org chart - starting specifically with t= he CX team, which had high costs, and where Atlan thought their unique proc= esses and flows wouldn=E2=80=99t be met well by external tools. =20 They started mapping =E2=80=9Ca day in the life,=E2=80=9D employee by emplo= yee, and evaluating the =E2=80=98jobs to be done=E2=80=99 to see which coul= d be most impacted by AI:=20 In the new world org, one (human) customer success manager would have sever= al (agent) employees, for instance, a call intelligence lead (to prep and f= ollow up), or a health intelligence lead (for reading product analytics, fe= eding intelligence back for the decision-making).=20 They found that the more specific they could make the agents, the better th= e agents did, ultimately becoming better than humans at doing the sub-task.= This resulted in a single customer service agent having up to nine agents = doing specific aspects of their job for them: The results for a company of just 400 were impressive. They now have more t= han 152 agents in their AI registry, and have had more than 4,000 agent run= s in just 5 weeks. Many processes were reduced from several hours to less t= han one, and experienced some capabilities that just weren=E2=80=99t even f= easible before (mid-call briefings on attendees by an agent, for instance) = are now joyfully in flow. Why Atlan Moved From Prompts to Agents=20 Atlan=E2=80=99s early initiatives used prompts with Notebook LLMs but evolv= ed to agents for even more impact. While prompts are simple to set up and g= reat for static tasks (summaries, rewrites, Q&A), they don=E2=80=99t captur= e memory or continuity across steps, and can=E2=80=99t autonomously decide = what to do next. The agents can reason, plan, and take actions across different tools/data s= ources (such as calendars, CRM, and product data), and iterate to achieve t= heir goals. This was useful for Atlan=E2=80=99s complex tasks like monitor= ing account health, drafting personalized success plans and triggering updates. Though the agents are more complex to design and govern, the results have b= een more than worth it for Atlan. =20 What was their tech stack?=20 Key Learnings from Atlan=E2=80=99s AI-Native Journey Pick the right problems =E2=80=93 Rethinking core, high-impact workflows fr= om first principles led to the biggest wins. Specialize your agents =E2=80=93 Narrow, task-specific agents consistently = outperform general ones. Context is king =E2=80=93 Data quality and tight feedback loops make or bre= ak effectiveness. Especially in customer-facing forums, the data must be ri= ght. Build cross-functional pods =E2=80=93 Pair a platform owner, an engineer, a= nd a workflow expert to move fast and effectively. Make it fun =E2=80=93 Contests, playful names, and community kept adoption = high and sticky. You need the team=E2=80=99s participation and feedback for= success. After introducing herself in Slack like a (human) employee, Atlan= =E2=80=99s Health Intelligence agent, Hermioni, was celebrated with a party= and a cake: =20 The irony?=20 While Prakulpa=E2=80=99s team is an impressive rocket of growth and AI depl= oyment, she=E2=80=99s measured about her AI optimism: =E2=80=9CI think all = of the companies that survive will have extensive AI agents - I don=E2=80= =99t know that it will be a competitive advantage. But at a high-growth sta= rtup, it=E2=80=99s hard to hire fast enough - so if we can do more with les= s, we can do more.=E2=80=9D=20 Have you started rethinking workflows for an AI-Native world yet? I=E2=80= =99d love to hear your stories. Carilu Dietrich is a former CMO, most notably the head of marketing who too= k Atlassian public. She currently advises CEOs and CMOs of high-growth tech= companies. Carilu helps leaders operationalize the chaos of scale, see aro= und corners, and improve marketing and company performance. Please subscribe to receive new posts and support my work. Unsubscribe https://substack.com/redirect/2/eyJlIjoiaHR0cHM6Ly93d3cuY2FyaWx= 1LmNvbS9hY3Rpb24vZGlzYWJsZV9lbWFpbD90b2tlbj1leUoxYzJWeVgybGtJam95TURrd01UYz= BNallzSW5CdmMzUmZhV1FpT2pFM05EQTBNamN4Tnl3aWFXRjBJam94TnpVNU1qUTJOamd6TENKb= GVIQWlPakUzT1RBM09ESTJPRE1zSW1semN5STZJbkIxWWkweE5UUTRNREk0SWl3aWMzVmlJam9p= WkdsellXSnNaVjlsYldGcGJDSjkucDdfRFJvZmpHODNKaFVmQ3JkR0wtVEdnY2E1cHRNVzhnVTF= aelVleXV6VSIsInAiOjE3NDA0MjcxNywicyI6MTU0ODAyOCwiZiI6dHJ1ZSwidSI6MjA5MDE3ND= I2LCJpYXQiOjE3NTkyNDY2ODMsImV4cCI6MjA3NDgyMjY4MywiaXNzIjoicHViLTAiLCJzdWIiO= iJsaW5rLXJlZGlyZWN0In0.7iVprFsmcc4hkkwGsIrNhU2S5CgSYyJBfd3t_MA12pM? --5ba85d8c17e17b21ec2f68599823247667e7978b459ab2535e98ccea49a3 Content-Type: text/html; charset="utf-8" Content-Transfer-Encoding: quoted-printable Case Study: Moving from = AI-First to AI-Native3D""
How Atlan Re-Engineered its Company, Processes, and People to becom= e AI-Native
͏     ­͏     ­͏=     ­͏     ­͏     = 73;͏     ­͏     ­͏   &#= 8199; ­͏     ­͏     ­͏ =     ­͏     ­͏     = 3;͏     ­͏     ­͏   = 199; ­͏     ­͏     ­͏ &= nbsp;   ­͏     ­͏     ­= ;͏     ­͏     ­͏   Q= 99; ­͏     ­͏     ­͏ &n= bsp;   ­͏     ­͏     ­= ͏     ­͏     ­͏   ̳= 9; ­͏     ­͏     ­͏ &nb= sp;   ­͏     ­͏     ­&= #847;     ­͏     ­͏    = ; ­͏     ­͏     ­͏ &nbs= p;   ­͏     ­͏     ­&#= 847;     ­͏     ­͏    = ­͏     ­͏     ­͏  = ;   ­͏     ­͏     ­= 47;     ­͏     ­͏     = ­͏     ­͏     ­͏  =   ­͏     ­͏     ­T= 7;     ­͏     ­͏     &= #173;͏     ­͏     ­͏   =   ­͏     ­͏     ­͏= ;     ­͏     ­͏     &#= 173;͏     ­͏     ­͏   &= #8199; ­͏     ­͏     ­͏=     ­͏     ­͏     = 73;͏     ­͏     ­͏   &#= 8199; ­͏     ­͏     ­͏ =     ­͏     ­͏     = 3;͏     ­͏     ­͏   = 199; ­͏     ­͏     ­͏ &= nbsp;   ­͏     ­͏     ­= ;͏     ­͏     ­͏   Q= 99; ­͏     ­͏     ­͏ &n= bsp;   ­͏     ­͏     ­= ͏     ­͏     ­͏   ̳= 9; ­͏     ­͏     ­͏ &nb= sp;   ­͏     ­͏     ­&= #847;     ­͏     ­͏    = ; ­͏     ­͏     ­͏ &nbs= p;   ­͏     ­͏     ­&#= 847;     ­͏     ­͏    = ­͏     ­͏     ­͏  = ;   ­͏     ­͏     ­= 47;     ­͏     ­͏     = ­͏     ­͏     ­͏  =   ­͏     ­͏     ­T= 7;     ­͏     ­͏     &= #173;͏     ­͏     ­͏   =   ­͏     ­͏     ­͏= ;     ­͏     ­͏     &#= 173;͏     ­͏     ­͏   &= #8199; ­͏     ­͏     ­͏=     ­͏     ­͏     = 73;͏     ­͏     ­͏   &#= 8199; ­͏     ­͏     ­͏ =     ­͏     ­͏     = 3;͏     ­͏     ­͏   = 199; ­͏     ­͏     ­͏ &= nbsp;   ­͏     ­͏     ­= ;͏     ­͏     ­͏   Q= 99; ­͏     ­͏     ­͏ &n= bsp;   ­͏     ­͏     ­= ͏     ­͏     ­͏   ̳= 9; ­͏     ­͏     ­͏ &nb= sp;   ­͏     ­͏     ­&= #847;     ­͏     ­͏    = ; ­͏     ­͏     ­͏ &nbs= p;   ­͏     ­͏     ­&#= 847;     ­͏     ­͏    = ­͏     ­͏     ­
<= table role=3D"presentation" width=3D"auto" border=3D"0" cellspacing=3D"0" c= ellpadding=3D"0">
Forwarded this email? Subscribe he= re for more

Case Study: Moving from AI-First to AI-Native

Ho= w Atlan Re-Engineered its Company, Processes, and People to become AI-Nativ= e

=
=
<= /tbody>
 
=
=
=3D""
=
= 3D""
3D""
<= tr>
READ IN APP
 

What is the difference between ‘AI-Forward’= ; or ‘AI First’ and truly ‘AI Native’? Can you get = native if you weren’t born that way?

At the 10= X CEO Conference two weeks back, Prukalpa ⚡ S= ankar, founder and co-CEO of Atlan, laid out a detailed roadmap and toolset= of their journey to becoming AI-Native that was impressive.

After substantial focus and push, this company of 400 has reimag= ined roles and jobs to be done, built more than 152 agents that made 4,000 = runs in just 5 weeks. They have massively sped up processes and unlocked ne= w responsiveness that wasn’t possible before, especially in customer = support. The company is AI-ifying across every role and toolset. I share he= r tech stack, process, and takeaways below.

For her:

<= blockquote style=3D"border-left: 4px solid #0052cc;margin: 20px 0;padding: = 0;">

'Being "AI First" means AI is considered= as the primary solution. But being "AI Native" means AI is fundamentalR= 12;integrated into the fabric of our thinking, operations, and processes. I= t means reimagining workflows from first principles in a world powered by A= I, rather than simply layering AI onto existing processes.”

Four-Part AI-Native Roadmap

Atlan to= ok a four-part journey on its sprint to AI-Native: Phase 1: AI Taskforce, P= hase 2: Shift the culture, Phase 3: Embed into hiring, Phase 4: Rethink the= org chart.

Phase 1: AI taskforce

Pruka= lpa’s first step, one many teams have taken, was to assemble an elite= team of people who had already shown passion and skill in AI. They were ta= sked with taking the most impactful company use-cases live, determining com= pany-wide tooling, and driving enablement.

Phase 2: Shif= t the culture

She then launched an AI Native mission to de= fine overall strategy across three long-term pillars:

Next came an AI prod= uctivity challenge to the company - a time-bound contest whe= re team members could win prizes and be celebrated for identifying where th= ey could leapfrog their productivity with AI. More than 300 people contribu= ted, and AI started entering workflows across the business.

Phase 3: Embed into hiring

Next, AI became part= of the hiring process. Atlan introduced hiring veto power for candidates w= ho didn’t have enough AI curiosity or AI skill. In the interviews, At= lan made it explicit that candidates were expected to use AI extensively, e= ven and especially in interview project rounds. They even asked candidates = to open up AI and prompt in front of the interviewer to see how they intera= ct and think using AI. Their guidance to the team on interviewing for AI sk= ills:

<= a class=3D"image-link" target=3D"_blank" href=3D"https://substack.com/redir= ect/20aced32-9504-4697-b220-a80a10a2ad5d?j=3DeyJ1IjoiM2dmeXZtIn0.xu76uFObqA= rDfP822j-jnN48_jCfgM3m0rbAsF0l24U" rel=3D"" style=3D"position: relative;fle= x-direction: column;align-items: center;padding: 0;width: auto;height: auto= ;border: none;text-decoration: none;display: block;margin: 0;">3D""
3D""

Why Atlan Moved From Prompts to Agents

Atlan’s ear= ly initiatives used prompts with Notebook LLMs but evolved to agents for ev= en more impact. While prompts are simple to set up and great for static tas= ks (summaries, rewrites, Q&A), they don’t capture memory or conti= nuity across steps, and can’t autonomously decide what to do next.

The agents can reason, plan, and take actions across di= fferent tools/data sources (such as calendars, CRM, and product data), and = iterate to achieve their goals. This was useful for Atlan’s complex = tasks like monitoring account health,
drafting personalized= success plans and triggering updates.

Though the ag= ents are more complex to design and govern, the results have been more than= worth it for Atlan.

What was their tech stack?

=
3D""

Key Learnings from At= lan’s AI-Native Journey

  1. Pick the right problems – Rethinking core, high-impact workflows from first principles led= to the biggest wins.

  2. Special= ize your agents – Narrow, task-specific agents consist= ently outperform general ones.

  3. Context is king – Data quality and tight feedback lo= ops make or break effectiveness. Especially in customer-facing forums, the = data must be right.

  4. Build cro= ss-functional pods – Pair a platform owner, an enginee= r, and a workflow expert to move fast and effectively.

  5. Make it fun – Contests, play= ful names, and community kept adoption high and sticky. You need the team&#= 8217;s participation and feedback for success. After introducing herself in= Slack like a (human) employee, Atlan’s Health Intelligence agent, He= rmioni, was celebrated with a party and a cake:

<= /td>= 3D""

The irony?

= While Prakulpa’s team is an impressive rocket of growth and AI deploy= ment, she’s measured about her AI optimism: “I think all of the= companies that survive will have extensive AI agents - I don’t know = that it will be a competitive advantage. But at a high-growth startup, it&#= 8217;s hard to hire fast enough - so if we can do more with less, we can do= more.”

Have you started rethinking workflows for an= AI-Native world yet? I’d love to hear your stories.

Cari= lu Dietrich is a former CMO, most notably the head of marketing who took At= lassian public. She currently advises CEOs and CMOs of high-growth tech com= panies. Carilu helps leaders operationalize the chaos of scale, see around = corners, and improve marketing and company performance.

Please subscribe to receive new posts and support my work.

<= div class=3D"subscribe-widget is-signed-up is-fully-subscribed" data-compon= ent-name=3D"SubscribeWidget" style=3D"margin: 0 0 1em;direction: ltr;font-s= ize: 16px;line-height: 26px;">
3D""Subscribed

<= /p>

 =
3D=Like=
= =
3D""Comment
3D""Restack
 

3D"Get3D"Start

3D""3D"" --5ba85d8c17e17b21ec2f68599823247667e7978b459ab2535e98ccea49a3--