# Joey Glyshaw

Joey Glyshaw shares practical guidance on using AI, integrations, and custom software to improve Amazon and ecommerce operations, including inventory forecasting, order fulfillment, and post-purchase support.

- Profile: https://b2binfluencer.marketing/creators/joey-glyshaw/
- LinkedIn: https://www.linkedin.com/in/joeglyshaw
- LinkedIn followers: 4,563
- Topics: artificial-intelligence, agentic-commerce
- Based on 25 dated posts, latest 2026-10-06

## Summary

Most useful for practical guidance on AI, integrations, and custom software in Amazon and ecommerce operations.

- Best for: Post-purchase, tracking, or fulfillment automation platforms seeking advice or co-produced content on AI-driven Amazon FBA operations.
- Not for: Subscription operations or warehouse-management software vendors seeking sponsored endorsements or paid placements.
- Check first: Ask for data behind outcome claims and LinkedIn audience analytics before contacting.

### Takeaways

- AI cannot call APIs on its own; MCP servers are the fix.
- A store is becoming a data feed an AI will either trust or skip.
- Own your operating software instead of renting a dozen SaaS tools.

## Main ideas

- **AI cannot call APIs on its own. MCP servers are the fix.**: Glyshaw returns to one technical gap: an API is how one piece of software talks to another, and language models are good at language but bad at APIs because every API has its own endpoints, logins, and data formats. When a model guesses, things break. An MCP server fills the gap: it tells the model which tools exist and how to use each one, makes the correct call, and hands back the answer. Their point: the API never went away; MCP is the layer that lets AI use it properly.
- **A store is becoming a data feed an AI will either trust or skip**: Their forward-looking argument: AI agents will complete a meaningful share of purchases without customers seeing a cart page, checkout, or abandoned session. The agent finds the product, confirms the price, applies a coupon, and buys. Product listings, they argue, will stop being sales pages aimed at human eyes and become machine-readable data: structured product data, clean API endpoints, and checkout flows an agent can drive.
- **Own your operating software instead of renting a dozen SaaS tools**: “SaaS is dead” recurs across several posts. The reasoning is consistent: brands rent a dozen tools that each do most of the job, none of which talk to each other, and fill the gaps by hand. Their alternative is one system shaped around how the business already runs, housed in the brand’s own copy, subdomain, and database, with an MCP so the brand’s own Claude or ChatGPT can query the data. They describe a standard base—dashboard, P&L, PPC, listings, inventory and forecasts, approvals, and role-based logins—then client-specific additions such as a Shopify section, TikTok Shop, a 3PL feed, or a creator tracker.
- **The AI gap is ownership, not tool access**: Glyshaw argues that companies are not under-tooled but under-structured. The missing piece is not another subscription; it is one person inside the business responsible for making AI work: mapping a workflow, finding the process that costs the most human hours, building the automation, and training the team to use it—or a fractional chief AI officer. Related posts cast an agent not as a platform feature but as a portable folder of markdown files holding identity, knowledge, instructions, and memory. The work then survives a platform change the way a website survives a browser change.
- **AI forecasting is a working-capital decision, not a dashboard purchase**: They compare spreadsheet forecasting with agent-run pipelines, framing the difference in financial terms rather than features: lower forecast error, fewer lost sales from stockouts, lower inventory levels at the same service level, and less analyst time. The mechanism is unglamorous: pull more than a year of sales history, layer in return rates, FBA levels, seasonal trends, and outside signals, then generate restock recommendations before anyone opens Seller Central.
- **The warehouse floor supplies the data that makes the AI useful**: Their most personal argument is that physical operations and AI building belong to the same job. Running a 3PL generates real SKU movement, carrier delay patterns, and return reasons that no synthetic training set reproduces. That data separates a useful system from an impressive demo. The reverse also holds in their telling: every repetitive task on the floor becomes a specification for the automation that should replace it.
- **Amazon operating discipline stands on its own, without the AI framing**: A strand of their writing focuses on seller operations. Profitable sellers, they argue, do fewer things with sharper discipline: pick margin before volume, register the brand before scaling ads, run PPC like a finance lead, stay in one or two categories, and treat cash flow—not the listing—as the real constraint. A companion post steers sellers toward no-code automation, recommending they start with one automation and measure time saved before expanding.
- **They teach automation through named tools and numbered steps**: Their how-to posts follow a clear method: define the manual pain, break the fix into numbered steps, name the software at each step, recommend a small test before a full rollout, and close with a checklist to request. The specific tool names make the instructions inspectable rather than promotional.

## Work to read

- [A day split between the 3PL floor and an AI agent stack](https://www.linkedin.com/posts/joeglyshaw_amazonfba-3pl-aiinfrastructure-activity-7511058000271273986-51PV)
- [SaaS is dead: build the operating software instead of renting it](https://www.linkedin.com/posts/joeglyshaw_saas-is-dead-again-stop-renting-12-activity-7510824419062185985-h-wA)
- [Replacing a $1,997-per-month warehouse system with something built in-house](https://www.linkedin.com/posts/joeglyshaw_i-was-paying-1997-mo-and-10-per-order-activity-7496559285150236672-2rrB)
- [A five-step post-purchase automation walkthrough](https://www.linkedin.com/posts/joeglyshaw_92-to-97-percent-forecasting-accuracy-that-activity-7456055525051404288-4WJa)
- [The agentic commerce prediction: AI buyers and checkout readiness](https://www.linkedin.com/posts/joeglyshaw_by-the-end-of-2026-a-meaningful-portion-activity-7467256423077314561-phFg)
- [What separates profitable Amazon FBA sellers in 2026](https://www.linkedin.com/posts/joeglyshaw_86-of-amazon-fba-sellers-say-they-are-profitable-activity-7445501844891549697-7cpA)
- [APIs versus MCP servers, explained plainly](https://www.linkedin.com/posts/joeglyshaw_apis-were-built-for-software-talking-to-software-activity-7511510177469112321-cCVR)
- [Turning handwritten sketches into explainer videos](https://www.linkedin.com/posts/joeglyshaw_remarkable-aiworkflow-aiagents-activity-7511483977782861825-iKpc)

## Brand relationships

- **Amazon** (mention)
- **Shopify** (mention)
- **AlgoFuse AI** (mention)
- **Anthropic** (mention)
- **OpenAI** (mention)

## Partnerships

The posts describe how people engage Joey Glyshaw commercially, but don’t present a sponsorship package. They offer free 30-minute AI audits as an entry point to paid build work, including rebuilding an existing software product for an audited business at a fraction of its previous cost. They also invite operators, agencies, and service businesses to a 30-minute discovery call with no pitch, and use comment-triggered lead magnets, a free group, and a beta list for tools they’re building themselves. For a brand, the practical route is a direct conversation: fit is decided on a call or audit, and work is scoped around the client’s own systems rather than offered from a public rate card. No rate card, deliverable schedule, or sponsorship terms appear in the posts; ask directly whether they accept sponsorships and what terms they offer.

## Alternatives

- Bernard Nader: Choose Bernard Nader when a campaign needs to reach Amazon sellers who suspect advertising is not their real problem. He works backwards from fees and unit economics instead of ad settings. One post takes a parent listing with five child SKUs and shows one child losing $4,078 on $29,519 in sales while another made $3,087 on $19,587. It then lists what to check before touching bids: $5,287.68 in FBA inbound defect fees inside a $42,164.76 fee total; whether Amazon has the right dimensions and weight; packaging, storage and pricing. The line that carries the argument: you cannot bid-optimise your way out of bad unit economics. A second post gives an unusually candid account of telling a client PPC could not fix the business even as CPC, ACoS, CTR and TACoS improved—and losing the account for saying so. For a platform claiming automation improves the numbers underneath the ad account, this is the strongest available version of the same case.
- Osaf Ali: Choose Osaf Ali when a campaign aimed at sellers facing risk rather than growth is needed. The posts deal with the failures that stop a seller from investing at all. One works through the document chain Amazon expects in an account health case now that the interview stage is gone: brand-to-distributor and distributor-to-seller letters of authorisation, warehouse agreements and invoices, distributor-to-prep-warehouse tracking, and a signed bill of lading from the prep warehouse to Amazon, with the point that Amazon may want to see how inventory moved, not only where it was bought. The other, from May 2026, explains the meltable-inventory problem, where sellers who shipped during the restricted window face removal fees, storage and return shipping, and recommends creating an FBM offer on the same ASIN fulfilled through Amazon MCF as often the cheaper route out, plus checking the meltable ASIN list before shipping grocery products between March and October. For a platform selling to sellers who have been burned by a policy change, that risk-management register opens a conversation a feature pitch cannot.
- Abhiraj Aggarwal: Choose Abhiraj Aggarwal when a campaign needs to address Amazon's own AI tooling and the marketplace economics around it. One post examines the Accelerate 2026 announcements—the Selling Partner plugin connecting Seller Central to Claude, Seller Assistant's persistent memory of a seller's pricing patterns and inventory cycles across sessions, and the canvas workspace for modelling a price change—and separates what changed, much faster access to your own data, from what did not: knowing what to do about it. It also flags that the connections are in beta and recommendations will need checking, a rare note of candour in an announcement write-up. The other post traces Amazon's margin enforcement: ASINs pushed into Direct Fulfillment when they miss contribution-margin targets, search suppression for items below profitability thresholds, and Gross Margin Guarantee pressure on vendors. Aggarwal draws on the client accounts his agency works on. For a platform competing with or integrating alongside Amazon-native assistants, he brings landscape analysis and caveats in the same piece.

---
AI-assisted summary of public posts. Audience demographics are not available. Engagement appears only where we checked recent posts. Method: https://b2binfluencer.marketing/methodology/