Role & team shape

Clint Borrill is Director of SEO at Balsam Brands. Based in Redwood City, 15 years at Balsam. Sits inside Marketing. Leads Balsam’s SEO organization — Chris Gimena manages the SEO team in Manila (strategists, analysts, content writers, project coordinators), and there’s a Manila team offsite the week of June 3 that Clint is planning for.

Day-to-day collaborators outside his own team:

• Swati Ayyar — Merch Ops — on PDP automation and content data.
• Mary Corrick — PIM Owner — on product-data quality and how it flows from PIM to the downstream channels.
• Joanne Alcayaga-Talha — Marketing Technology Manager, BHUS — owns the Oculate/Feedonomics integration work. Clint refers to her as “Joni” in conversation.
• Oculate — external PDP automation partner.

Weekly time breakdown

Clint didn’t break his week into a numeric split. The call painted a picture of time spread across SEO strategy, the ongoing Oculate/PDP automation project, feed architecture (MuleSoft + Feedonomics), and hands-on AI experimentation — he’s currently building a product-description skill in Claude’s Skill Builder on the side. Meeting with the growth team tomorrow on a specific Amazon-copy-in-house thread.

Annual volume / scope

Clint’s lens covers the full product-content surface across every channel Balsam sells through:

• All English-speaking geos (universal brand tone + local nuances — Germans want details, Americans want flowery language, etc.).
• Every product channel: balsamhill.com, Amazon, Target, Meta (shops/ads), Google Merchant Center, ChatGPT, Perplexity, and other LLM-era feeds as they come online.
• All product categories and families. Re-architecting SEO from parent-level canonical tags (one page per category) to child-level SKU pages so agent search can match on granular attributes.
• The full PIF → PIM field set on which all of the above depends.

Process (his workflow)

The spine of Clint’s world is the data pipeline:

1. PIF (merchant-entered) — merchants key product info into the PIF Excel.
2. PIM (ContentServ) — product info lands in PIM. Brand copy team writes descriptions; Clint’s team layers in meta title, meta description, slugs, SEO attributes. Canonical tags set here.
3. MuleSoft — configures what data flows out from PIM.
4. Feedonomics — determines which data elements go to which channels, and handles enrichment. Bidirectional: can enrich outbound feeds or enrich and push back into source systems. Clint flagged the second direction as underused.
5. Channel syndication — feeds go to Google Merchant Center, Amazon, Target, Meta, ChatGPT, Perplexity, and each channel’s own requirements-set.

In parallel, the Oculate/PDP automation project:

• Goal: auto-generate PDP content across English-speaking geos from PIM data, with universal brand tone + local nuance.
• Today’s state: over-specified. They pass Oculate every trademark rule, every height/light/foliage content requirement. “Really, really challenging.” Making it more complex than it needs to be.
• Forward state: a coming PDP redesign simplifies content to an inspiration/intro + use-case block at parent + family level, with meta title/description regenerated from there. No more child-level content requirements on the PDP itself.

Canonical strategy shift (in-flight):

• Parent-level was right for the Google era (one strong page per category).
• Agent-era search will match granular attributes (“I live in an apartment with 10-foot ceilings — what are good options?” → infer a 6.5–7.5 ft tree). That needs child-level pages with clean, complete, accurate SKU data — which doesn’t exist today.

Current AI usage

• Claude desktop app — his primary AI surface. “Fundamentally primarily with the desktop app.”
• Claude Code — dabbled in the terminal; not a heavy user yet. This was the first time he’d seen Claude Code driving a repo end-to-end.
• Claude Skill Builder — actively building a product-description skill. Interview-based: feeds in input files, context files, expected output, column/format definitions; gets a markdown skill file ready to commit. Thinks of skills as portable instructions across Claude, Cursor, etc. — platform-agnostic.
• ChatGPT — used across Balsam but flagged as problematic without guardrails. Priscilla’s GPT image-gen output (“making trains with two heads”) is the example he keeps returning to for why brand-grounded AI matters.

Blockers / pain points

• Child-level SKU data is incomplete and inaccurate. “So this data today is not very accurate when you actually look at the individual one. The more data, the better, I think is the bottom line for SEO. And accuracy.”
• Oculate partnership is struggling under complexity. The over-specified current PDP structure forces the automation to carry every edge case. The coming PDP redesign should reset this.
• Platform sprawl with no upstream source-of-truth strategy. Amazon, Target, Meta, Google, ChatGPT, Perplexity each want different descriptions/attributes. Today this lives in Feedonomics config and MuleSoft mappings — a downstream feed-rule problem rather than an upstream PIM architecture.
• Individual-level AI use leaves no org leverage. “Basically what you’re doing is given tools and people are doing their own thing everywhere… The context window is very finance [narrow].” Every person’s learning stays in their own ChatGPT session.
• Internal tool discovery is broken. DevOps has built tools; Marketing doesn’t know about them. Balsam’s internal MCP registry has zero MCP servers registered. Skills are being built but not shared.
• Brand guardrails don’t exist in AI tools. Priscilla’s spec-sheet GPT output drifts off-brand because nothing grounds it in Balsam’s aesthetic or past SKUs.

Wish-list / what he’d do with more time

• Bring Amazon product descriptions in-house. “Last year, they pay their marketing agency to write all our Amazon descriptions and stuff, and I’m like — we could do this ourselves.” Meeting with growth team tomorrow to scope. Believes it could be stood up “in a few hours” with the skill he’s already building.
• Per-platform source-of-truth fields in PIM. Add Amazon description, Target description, Meta description, etc. as first-class PIM fields. Flag a SKU as Amazon-allocated → auto-syndicate via Feedonomics. Replaces the current “all platforms share one description and Feedonomics munges it” pattern.
• Centralized Balsam product-image library AI tools can consume. So Priscilla’s GPT, or any other team’s, generates on-brand imagery from Balsam’s actual past SKUs instead of inventing two-headed trains.
• Competitive analysis at scale. Pointed at Alex’s Reddit work as the shape: “your Reddit thing is kind of a perfect example of how we can do totally new things because AI and because our data is organized.” Wants Priscilla out of Pinterest and into something that actually organizes competitor product ideas.
• Real-time compliance across feed channels. As Google algorithm updates, ChatGPT/Perplexity feed requirements shift — have the org able to validate in real time that product data meets each channel’s requirements.

Cross-functional gaps he raised

• SEO ↔ PD. Priscilla needs guardrailed AI for vendor spec generation and an image library that stays on-brand. Clint’s team has the strongest view of what granular data agent-era search will demand — and Priscilla’s team is upstream of it.
• SEO ↔ Data quality. Mary (PIM) and Swati (Merch Ops) own data that Clint’s team consumes downstream. Clint wants them looped into what agent-era search and per-platform syndication need, so enrichment isn’t just “complete the blanks” but “complete the blanks the downstream consumer actually needs.”
• SEO ↔ Growth. The growth team pays the marketing agency for Amazon descriptions Clint believes can be brought in-house. Meeting with growth tomorrow to align.
• DevOps ↔ everyone else. Internal tooling exists but there’s no discovery mechanism. “Zero MCP servers registered” in the internal registry. Skills and tools stay invisible to the functions that would use them.

Named colleagues he relies on or mentioned

• Swati Ayyar — Merch Ops. Collaborator on PDP automation; downstream data owner.
• Mary Corrick — PIM Owner. Collaborator on product-data quality and downstream consumption.
• Joanne Alcayaga-Talha — Marketing Technology Manager, BHUS. Owns the Oculate/Feedonomics integration. Clint calls her “Joni.”
• Priscilla Spicer — Product Development. Key user who needs guardrailed AI and a product-image library.
• Chris Gimena — SEO Manager, Manila. Runs Clint’s Manila team; offsite week of June 3.
• Oculate — external PDP automation partner. Relationship under strain from over-specified current requirements.
• Unnamed growth-team counterpart — scheduled 1:1 the day after this call to scope Amazon-copy-in-house.

Candidate ideas extracted from the call

Ranked by signal strength. Top 4 filed as GitHub issues this run.

1. Amazon product-description auto-drafter. He has a growth-team meeting booked, he’s already building the skill, the scope replaces paid agency work.
2. Per-platform product-data source of truth in PIM. The biggest architectural shift he proposed. Asked directly: “Would we want the source of truth for Amazon to be in our world? Do we want an Amazon description field within pim?”
3. Central Balsam product-image library for AI consumption. Complements Priscilla’s brand-grounded image generator (#34) — hers is the tool, this is the data layer underneath.
4. Child-level SKU data enrichment for agent-era search. Needed to support his canonical strategy shift from parent-level to child-level.
5. Competitive product-assortment scan — likely overlaps with Clint’s existing Reddit work; holding this run.
6. Internal tool/MCP registry for cross-function discoverability — CoE operating idea, not a cohort work product; holding this run.

Status

In conversation. Not confirmed. He gave multiple substantive yeses across the call — “Yeah, all in”; “this has been great”; “I’m super excited about those sort of stuff”; “looking forward to it” — and showed no hesitation about time commitment. Holding at in-conversation pending Claire sign-off per the Cohort 1 roster-gating rule: Kate and Mary are foundational, everyone else needs sponsor approval before moving to confirmed.

Notable quotes

• “We’re not going to have a website in 10 years… everything’s going to be through or managed [by] an agent… we’re going to become increasingly reliant on data feeds — one for Google Merchant Center, one for ChatGPT, one for perplexity, etc.”
• “Would we want the source of truth for Amazon to be in our world? Do we want an Amazon description field within pim?”
• “Last year, they pay their marketing agency to write all our Amazon descriptions and stuff, and I’m like — we could do this ourselves.”
• “This data today is not very accurate when you actually look at the individual one.”
• “Basically what you’re doing is given tools and people are doing their own thing everywhere.”
• “Your idea with the Reddit thing is kind of a perfect example of how we can do totally new things because AI and because our data is organized.”

Transcript reference

Goose Group Swerk doc: de390fb6-1ea0-4f1f-a8f9-3421df636c07 — not linked from any rendered page.

Meeting date: 2026-04-20.