Skip to content
Ameet Palkar
Contact

Case study · 5 min read

How we made the organization ~25% leaner, 15–20% faster, and more competitive, without a single layoff.

same throughput
~25% leaner org
average delivery time
15–20% faster
proof-of-concept turnaround
3 weeks → 5 days
estimate turnaround
7 days → 1 day

The moment

By 2024, the ground under design agencies had shifted. Clients who once accepted three-week proof-of-concept timelines were being courted by competitors promising days. Inbound lead volume across the industry was collapsing, and the quality of leads that we were getting was poorer by the day. Every pitch had more agencies in it, most of them cheaper than us.

I was Design Director at Lollypop Design (a Terralogic company), one of India's largest design agencies: a multi-hundred-person design organization across four studios in Mumbai, Bengaluru, Hyderabad, and Chennai. The existential question wasn't whether AI would change agency work. It was whether we'd re-architect ourselves deliberately, or be re-architected by the market.

We were now expected to lead that transition. This is what we did.

The diagnosis: we were paying senior designers to do grunt work

Before touching a single tool, I mapped where designer time actually went across a typical engagement. The pattern was consistent and uncomfortable:

  • Discovery was dominated by mechanical work: drafting question banks, transcribing stakeholder interviews, collating and formatting findings into reports.
  • Design production meant hand-building wireframes, design systems, and prototypes in Figma, then packaging them into static PDFs that took hours to assemble.
  • Research was worse. Our researchers (psychology-trained, brilliant at insight, not technical) spent the majority of their time reading transcripts and wrangling data before analysis could even begin.
  • Presales was the bottleneck the sales team complained about most: estimates took 6–7 days, POCs took 2–3 weeks, and both pulled large amounts of time off designers.

The insight that shaped everything: AI's value to an agency isn't making designers faster at design. It's removing the non-design work that surrounds design, so the hours we bill are hours of actual design thinking.

Where designer time goesBefore and after comparison across four stages: discovery, research, design production and presales. Before, non-design work takes most of the time in every stage: transcribing, collating and formatting in discovery; reading transcripts and wrangling data in research; hand-building wireframes and static PDFs in design production; estimates in a week and proofs-of-concept in three weeks in presales. After, design thinking takes most of the time: interrogating the synthesis, interpreting and pressure-testing research, design judgment and interactive builds, and estimates in a day with proofs-of-concept in five days. Proportions are illustrative.Where designer time goesNon-design workDesign thinkingDiscoveryBeforeTranscribing, collating, formattingAfterInterrogating the synthesisResearchBeforeReading transcripts, wrangling dataAfterInterpreting and pressure-testingDesign productionBeforeHand-building wireframes, static PDFsAfterDesign judgment, interactive buildsPresalesBeforeEstimates in a week,POCs in three weeksAfterEstimates in a day, POCs in five daysProportions are illustrative.
AI took on the work around design, so designers' hours went to the design itself.

The intervention

1. Re-architecting delivery, end to end

We rebuilt the design process from discovery through delivery around a simple division of labor: AI does the grunt work, designers do the thinking.

Discovery interviews are now transcribed and synthesized automatically; designers spend their time interrogating the synthesis, not producing it. Reports ship as interactive HTML experiences instead of static decks, generated in a fraction of the time and, as it turned out, dramatically better received by clients. Prototypes moved from static Figma flows to interactive builds clients can actually touch.

For the research practice, we built workflows that ingest interview recordings and quantitative datasets and return structured analysis the researchers then interpret and pressure-test. A team that had been drowning in transcript-reading got its analytical time back.

2. Making it stick: the enablement system

Tools don't transform organizations; adoption systems do. I've spent over a decade in learning and organizational development before design, and this is where that background did the heavy lifting.

I formed a small cross-studio pilot team: hand-picked practitioners who tested workflows on live projects, documented what worked, and codified it into playbooks the rest of the organization could follow. I ran open showcase sessions across all four cities, demonstrating real before/after outcomes from real projects: time saved, output quality, client reactions. Adoption spread because people saw evidence, not because they were told to comply.

I also built the business case for the tooling investment itself (cost against projected value) and secured leadership approval for shared AI workstations across studios. The estimation tooling we built on top of them is now standard practice in every location, and the pilot team runs the enablement cycle without me in the room.

3. Rebuilding the presales engine

I moved into presales deliberately, because that's where the market pressure was sharpest, and where design-side slowness was costing us deals.

  • Estimates: from about a week to a day. Automated estimation tooling, reviewed by leads instead of built from scratch by them.
  • POCs: from ~3 weeks to under 5 days. Interactive prototypes replaced static Figma decks. Clients open a link and experience the thinking instead of paging through a PDF.
  • Conversion held as the market shrank. Lead volume halved; our monthly closures didn't. We kept winning against lower-priced competitors, on the strength of speed and the quality of what prospects experienced in the first meeting.

I also personally closed a series of new accounts across two quarters (at premium pricing, in a contracting market), which mattered less as revenue than as proof that the new engine worked at the sharpest end of the funnel.

The organizational bet

How the transformation unfoldedA timeline of seven phases, in order: the mandate, where leadership asks me to lead the AI transition; the diagnosis, mapping where designer time actually goes; the pilot team, practitioners testing workflows on live projects; playbooks, where what works gets written down; showcases, real before and after results shown in all four studios; the presales rebuild, estimates and proofs-of-concept re-engineered; and the structural decision, attrition not backfilled. It ends in the outcome: ~25% leaner org, same throughput, and 15–20% faster delivery.The mandateLeadership asks me to lead the AItransitionThe diagnosisMap where designer time actually goesThe pilot teamPractitioners test workflows on liveprojectsPlaybooksWhat works gets written downShowcasesReal before/after results shown inall four studiosThe presales rebuildEstimates and POCs re-engineeredThe structural decisionAttrition not backfilled~25% leaner orgsame throughput15–20% fasterdelivery
From mandate to structural change: adoption ran as an organizational change programme, not a tool rollout.

Here's the decision that turned process improvement into structural change: as natural attrition occurred, we didn't backfill.

The organization became roughly 25% leaner over about two years, while sustaining the same project throughput and cutting average delivery time 15–20%. No layoffs. No death-march overtime. The efficiency was real, so the smaller org genuinely could carry the load.

And the quality question (the one every agency leader asks about AI) answered itself: during this period the studio won the DNA Paris Design Award, India's Best Design Award, and the International Design Award, including projects I directly led. Work produced by the leaner, AI-augmented organization was the most internationally recognized work in the studio's recent history.

What I'd tell another agency leader

  1. Map where time goes before you buy anything. Our biggest gains came from discovery, reporting, and presales, not from "AI design tools."
  2. Adoption is an OD problem, not a tooling problem. Pilots, playbooks, showcases, and internal champions moved us further than any license purchase.
  3. Efficiency without a structural decision is just slack. The courage move was letting the org get smaller as it got faster.
  4. Speed is a client experience. A five-day interactive POC doesn't just cost less; it tells the prospect what working with you will feel like. It became our best sales argument.
  5. Protect the thinking. Every hour we automated was reinvested in design judgment. That's why quality went up, not down, and why the awards came during the transition, not despite it.

I'm a design leader with 20+ years across organizational development and product design. I'm currently Design Director at Lollypop Design, where I lead design, AI transformation and design-led presales across four studios.