# Paul Deraval

Paul Deraval writes about applying AI agents to sales and business operations, with a focus on using company data to guide recommendations and deploying agents safely.

- Profile: https://b2binfluencer.marketing/creators/paul-deraval/
- LinkedIn: https://www.linkedin.com/in/paul-deraval
- LinkedIn followers: 5,052
- Topics: artificial-intelligence, ai-agents, enterprise-ai
- Based on 15 dated posts, latest 2026-09-11

## Summary

Paul Deraval is most useful for concrete examples of AI agents in sales and business operations.

- Best for: A B2B SaaS media analytics campaign for explainable agent coaching and governed skills.
- Not for: A consumer brand awareness campaign seeking broad reach and lifestyle content.
- Check first: Confirm audience composition source and recent results for comparable topics.

### Takeaways

- An internal NinjaCat sales agent pairs follow-up suggestions with sales principles and account context.
- An agent found a reported $750,000 retail order opportunity.
- Enterprise agents need tightly scoped, composable skills, not unrestricted access.

## Main ideas

- **Agents should explain why they recommend something, not just issue the action**: Deraval treats explanation as a design decision that costs almost nothing and changes the result. A black-box agent that hands over an answer can erode users' skills, a risk they call task amnesia. An agent that justifies each recommendation like a coach turns the interaction into a training loop: users move faster and absorb the reasoning behind the move. In their framing, the outcome is a choice organizations make, not a fixed consequence of the technology.
- **The durable asset is the data layer, not the model**: Deraval argues that two years of debate about which model wins misses where value accumulates. An agent can replicate a software feature quickly, but not years of account history, permission structures, workflow audit trails, and institutional memory. They expect software to split into systems of record that agents query and interfaces that agents bypass entirely. Companies that offer headless, consumption-based access—so agents read the record instead of people navigating a screen—come out strongest.
- **Enterprise agents need tightly scoped, composable skills**: Deraval separates what is technically possible from what an enterprise can responsibly run. Unrestricted computer access produces impressive demos and unacceptable risks: leaked credentials, financial liability, damaged customer trust, and regulatory exposure. They propose small, governed capabilities that can be chained together, so an agent that reads email cannot also send mail to anyone. Their summary: big vision, tiny iteration.
- **Judge agents by the business action they trigger**: Deraval frames agent value in commercial outcomes rather than speed or novelty, treating trust as the bottleneck between insight and action. Their recurring question: will someone act on what an agent finds?
- **Personalizing reporting at scale is the open problem**: Deraval contrasts two reporting worlds. Agencies serving thousands of small-business clients are pushed toward templated, cookie-cutter reports. Teams serving big brands have room to tune the narrative, which is why their reporting often looks better. They see scalable personalization—preserving narrative quality while serving many accounts—as the interesting question, not a solved problem.

## Work to read

- [SDR Sidekick: an agent that explains its recommendations](https://www.linkedin.com/posts/paul-deraval_theres-a-few-articles-making-the-rounds-activity-7497654703762132992-a1D8)
- [The data layer is the durable asset, not the model](https://www.linkedin.com/posts/paul-deraval_the-ai-software-debate-is-missing-something-activity-7500260007536533506-hfJo)
- [Governed skills: why enterprises cannot copy consumer agent setups](https://www.linkedin.com/posts/paul-deraval_openclaw-aka-clawdbot-showed-us-whats-activity-7434612301892190208-qD1A)
- [A proof of concept that found $750K in new revenue](https://www.linkedin.com/posts/paul-deraval_750k-in-new-revenue-from-one-insight-uncovered-activity-7429903523942707200-6017)
- [Personalized reporting at scale is the unresolved problem](https://www.linkedin.com/posts/paul-deraval_reporting-looks-different-depending-on-who-activity-7427003414410104832-M6AU)
- [Live agent builds as customer proof at a media-industry event](https://www.linkedin.com/posts/paul-deraval_winding-down-the-last-day-of-lead-to-cash-bash-activity-7395173366988476416-hH6_)
- [Why software economics are shifting toward intelligent action at scale](https://www.linkedin.com/posts/paul-deraval_smb-software-was-already-in-a-race-to-zero-activity-7473736397347110912-YtOb)

## Brand relationships

- **NinjaCat** (employment)

## Partnerships

Paul Deraval's posts document their work at NinjaCat building and deploying AI agents: an internal sales-development agent that explains the reasoning behind each recommendation, a customer pilot in home equipment retail that surfaced $750,000 in new orders before a contract was signed, and a conference session with media-company operations leaders on removing friction from reporting and fulfillment. The posts offer practitioner commentary and company case examples, but no sponsorship terms, rate card, media kit, or partnership inquiry route. Brands should ask directly about available formats, deliverables, pricing, approval process, and how to get in touch.

## Alternatives

- Akhilesh Perla: Choose Perla for a campaign aimed at enterprise architects evaluating Salesforce and MuleSoft. In two posts, Perla presents Salesforce as an expanding enterprise platform, arguing that data governance, integration, and agent capabilities must work together for AI to operate across business systems. The Salesforce–Informatica platform post makes the case through specific platform layers; a later post on Salesforce’s changing role speaks directly to architects and IT leaders. Perla is a stronger fit than Deraval for a campaign about platform architecture and integration, rather than how an agent can coach a seller through its recommendations.
- Ahikam Kaufman: Choose Kaufman when the audience is finance leaders weighing the costs and risks of putting AI into financial workflows. Kaufman connects agent usage—model choice, token consumption, and controls—to the broader question of whether a finance team can adapt if its systems or needs change. The post on AI costs and controls in finance lays out the operating considerations; the post on changing direction asks buyers to consider dependencies, portability, and system replacement. Kaufman is a better fit than Deraval when the campaign’s central question is financial governance or flexibility, not the quality of an agent’s recommendations.
- Mark Bidgood: Choose Bidgood for a consumer-goods campaign about data definitions and alignment across functions. In the post on product definitions in CPG, Bidgood shows how different definitions across marketing, supply chain, finance, and retail can lead AI to optimize against inconsistent signals. A second post on trusted data makes the related point that connected systems are not enough if the business does not agree on what its data means. Bidgood is a better fit than Deraval when the buyer’s first challenge is aligning business definitions before deploying AI, especially in CPG or retail media.

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