# Joseph Paul

Joseph Paul is a MongoDB Customer Success Manager who works on agentic AI and agent memory, including building agents for MongoDB’s go-to-market teams.

- Profile: https://b2binfluencer.marketing/creators/joseph-paul/
- LinkedIn: https://www.linkedin.com/in/josephfpaul
- LinkedIn followers: 1,530
- Topics: artificial-intelligence, sales, customer-success, ai-agents, go-to-market
- Based on 11 dated posts, latest 2026-10-03

## Summary

Joseph Paul is most useful for accessible technical explainers on agentic memory and practical AI work.

- Best for: Developer-tools or AI-infrastructure brands wanting educational content on agentic memory and technical explainers.
- Not for: Consumer or enterprise brands seeking paid high-reach sponsored placements with guaranteed impressions or a media kit.
- Check first: Ask for per-post LinkedIn analytics and sponsorship terms before budgeting a paid placement.

### Takeaways

- Agent memory is eight problems, not one, with layers from working to shared memory.
- Agents built for MongoDB’s go-to-market organization saved hundreds or thousands of hours.
- Hands-on skill badges reinforce application-design and performance practices for MongoDB teams.

## Main ideas

- **Agent memory is eight problems, not one**: Joseph Paul argues that agentic memory is treated as a single feature when it is really eight interacting layers: working, episodic, semantic, procedural, entity, taxonomic, long-term, and shared. Treat the memory engine as one unit, they say, and the system will forget, confuse, or hallucinate context. Each layer has a specific job. Working memory is the scratchpad for the current conversation and tool results; it dies when the session ends. Episodic memory records what happened, why, and the outcome. Semantic memory holds distilled facts. Procedural memory captures learned workflows. Entity memory tracks people, companies, and their relationships. Taxonomic memory organizes how knowledge is classified. Long-term memory handles persistence, compression, and decay. Shared memory lets one agent's knowledge reach another. Their complaint is that most frameworks ship working memory and maybe a basic version of semantic memory, leaving the other six for developers to build—which is why persistent memory remains hard.
- **Build agents whose value shows up as hours saved**: Joseph Paul's test for agent work is operational: agents should remove real work from a go-to-market organization. They say that, over two years at MongoDB, they specialized in agents and agentic memory and built agents that saved hundreds or thousands of hours across the GTM org. A nomination followed to help shape the AI tools the organization uses for productivity. This is a practitioner's priority, not a research agenda. It sits alongside the technical framework: the layers explain the engineering; the saved hours explain why anyone should care.
- **Training is risk reduction, not an afterthought**: Joseph Paul argues that intentional training is the best way to make applications scale efficiently, framing hands-on instruction as a way to avoid costly design and performance mistakes rather than a box to tick. They recommend MongoDB Skill Badges to reinforce best practices, pointing to MongoDB's own Customer Success organization, which spends time earning badges. The post ends with an offer to set up a Skill Badge Day for any team that wants one.
- **Sell technology by mapping a customer's workflow to a number**: Before the AI work, Joseph Paul's software-selling method was to analyze a customer's workflow, quantify bottlenecks, and calculate return on investment so a manager could justify the purchase. The other half of that method was building relationships: cultivate champions among midlevel managers and influential engineers, then give them a stage. They hosted yearly conferences where those champions could present their collaborations and lessons learned to a global peer group. The same instinct—turning a technical claim into a figure a leader can approve—later shows up in descriptions of agent work through saved hours.
- **The framework is presented inside one company's products**: Joseph Paul's technical and educational writing is anchored in MongoDB. The eight-layer memory post promises a follow-up mapping MongoDB products to each layer; the training post recommends MongoDB Skill Badges; the two-year reflection mentions MongoDB announcements at a .local event. Readers should treat the product context as part of the position, not a detail.

## Work to read

- [The eight layers of agentic memory](https://www.linkedin.com/posts/josephfpaul_mongodb-agenticmemory-aiengineering-activity-7498087869568188417-Ce4L)
- [Two years at MongoDB: building agents for go-to-market teams](https://www.linkedin.com/posts/josephfpaul_i-recently-hit-2-years-at-mongodb-and-feeling-activity-7510475170294210560-IV7V)
- [Why hands-on skill badges matter for MongoDB teams](https://www.linkedin.com/posts/josephfpaul_whether-youre-brand-new-to-mongodb-or-you-activity-7405352524142084096-mX69)
- [First day as a MongoDB Customer Success Manager](https://www.linkedin.com/posts/josephfpaul_thrilled-to-announce-that-today-is-my-first-activity-7238943584786247681-9X9i)
- [Layoff at NI and a job-search post that doubles as a sales record](https://www.linkedin.com/posts/josephfpaul_long-post-dear-family-friends-and-network-activity-7212555895438729216-mAdk)

## Brand relationships

- **MongoDB** (employment)
- **NI** (employment)

## Partnerships

Joseph Paul's public posts cover their customer-success work at MongoDB, a MongoDB Skill Badge training day, an eight-layer framework for agentic memory in AI systems, and two career milestones. They do not describe a sponsorship offer, rate card, media kit, partnership program, paid brand collaboration, sponsored format, or official brand inquiry channel. Ask directly about availability and terms before budgeting a paid placement. Their content is tied to their employer's product area, so any commercial arrangement would sit alongside their MongoDB role rather than operate as a separate independent media business.

## Alternatives

- Terry Miles: Terry Miles is the better choice when a sponsor needs a teaching-led explainer on context engineering and memory for procurement, IT, and other enterprise stakeholders rather than application developers. The posts offer a practical sequence: start with a narrow, valuable question and run toward complex legacy data (23 July 2025), then define context engineering and short- and long-term memory in instructional-design terms, including confidentiality and procedural versus episodic memory (24 July 2025). For a webinar that helps buyers scope an AI initiative and bring IT and security allies in early, Terry Miles offers a more governance-oriented teaching frame than Joseph Paul's eight-layer build taxonomy.
- Conor Brennan-Burke: Conor Brennan-Burke is the better choice when a sponsor wants a founder-led, product-category argument about why agent context fails and what a context layer should do. The 25 March 2026 post argues that most business context is unstructured and lives across Slack, email, docs, and CRM; the 30 June 2026 post uses the blind-mice parable to explain that more connectors give an agent more fragments without resolution. For technical buyers already convinced memory matters and now evaluating architectures, that failure narrative and product point of view can be sharper than a neutral eight-layer explainer.

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