AI Sales Desk Notebook with Unified Context

AEs stopped juggling 10 tabs—they live in the notebook now.

Context

Our account executives (AEs) were spending an inordinate amount of time juggling multiple tools, a process that severely hampered their efficiency and deal velocity. Imagine a typical day: flipping between Attio for CRM data, Notion for internal playbooks, Google Docs for collaborative notes, and endless Slack threads for team communications—all while trying to close complex deals. Discovery call notes would reside in one document, security responses from specific clients in another, and critical finance approvals would be buried deep within email chains. This constant context switching wasn't just annoying; it actively destroyed deal momentum, led to overlooked details, and made the onboarding of new sales representatives an excruciatingly painful, weeks-long endeavor.

The core problem was a fragmented universe of information. There was no single source of truth for all deal-related context, leading to repetitive information gathering and a high cognitive load for our sales team.

To address this critical challenge, I developed an innovative "AI Sales Desk Notebook." This isn't just a static document; it's a dynamic, intelligent canvas where AI seamlessly assembles real-time pipeline data, comprehensive customer history, and proactive recommended actions. The notebook functions as a living document: AEs retain full editorial control, allowing them to add their insights and adapt content as needed. Crucially, though, the underlying system automatically keeps the integrated data fresh and enforces critical sales processes, such as mandatory fields or approval workflows. This has transformed our sales operations, dramatically improving deal velocity, streamlining onboarding, and providing a unified, intelligent workspace for our sales team.

Stack I leaned on

• Notion with synced blocks for live data: Notion serves as the flexible canvas for our sales desk notebook. We leverage its powerful synced blocks feature to pull in live, dynamic data from various sources directly into the notebook, ensuring reps always have the most current information.
• Attio API to pull pipeline + notes: Attio, our CRM, is integrated via its API to feed real-time pipeline data, deal stages, and historical notes directly into the Notion notebook. This eliminates manual data entry and ensures consistency between the CRM and the sales desk.
• OpenAI for summaries and suggestions: OpenAI's language models are at the heart of the AI capabilities. We use it to generate concise summaries of call transcripts, draft follow-up emails, and provide intelligent suggestions for next steps or potential blockers, significantly reducing the AE's cognitive load.
• Resend to trigger follow-ups directly: Resend is integrated to enable AEs to trigger personalized follow-up emails directly from within the Notion notebook. This streamlines post-call actions, ensuring timely communication with prospects and customers.

The Failure Modes We Needed to Fix

1. Fragmented notes: discovery summaries, MEDDIC/CHAMP fields, and technical blockers never existed in the same place, so handoffs were messy.
2. Lost follow-ups: reps left calls with action items but no automation to draft the recap or schedule next steps.
3. Compliance gaps: selling into regulated industries required logging who accessed which doc and when; Notion alone wasn’t enough.
4. Onboarding lag: new hires spent two weeks learning “which tab holds what” instead of learning the product.

Every requirement for the notebook traced back to eliminating one of these failure modes.

Notebook Architecture

| Layer | Purpose | Notes | |-------|---------|-------| | Data sync | Pulls realtime CRM data | Attio webhooks push updates into a Supabase edge function that writes to Notion synced blocks | | Notebook template | Standardizes deal documentation | A Notion template with sections for context, stakeholders, risks, experiments, and mutual action plan | | AI services | Summaries + drafting | OpenAI GPT-4o mini with structured prompts referencing the latest synced data | | Automation hooks | Follow-ups, task creation | Resend for personalized emails, Linear for engineering asks, Slack for exec updates | | Observability | Logging + analytics | BigQuery stores every AI suggestion, human edit, and send event for audits |

The template is duplicated per deal and bound to the CRM opportunity ID. When the status changes in Attio, the notebook receives a webhook and updates relevant sections instantly.

Playbook for Building the Desk

1. Design the golden template: I mapped our best-performing deals and encoded every section we wished we’d captured earlier—business context, pains, politics, land-and-expand strategy, mutual success plan.
2. Wire data syncs: Supabase functions listen to Attio webhooks and update Notion synced blocks (ARR, stage, health score, procurement notes). No one touches these blocks manually.
3. Layer AI copilots: Buttons labeled “Summarize last 3 calls,” “Draft recap,” and “Suggest next experiment” call our AI service. Responses land in a staging block until the rep accepts/edits.
4. Embed guardrails: Sensitive snippets (pricing, security promises) are templated. If AI drafts text referencing these areas, it must cite the approved language and link back to the source doc.
5. Integrate actions: Reps trigger Resend follow-ups or create Attio tasks from inside the notebook. Context (call notes, next steps) is injected automatically, but humans approve before sending.
6. Instrument feedback loops: Each notebook has a “feedback” synced block. Reps tag issues (data missing, tone off, automation bug). A nightly job aggregates tags and opens GitHub issues.

Key Principles of the AI Sales Desk Notebook

• Unified context: Consolidate all relevant deal information, customer history, and internal resources into a single, easily accessible canvas.
• AI-powered assistance: Leverage AI to automate repetitive tasks, summarize information, and provide intelligent suggestions, augmenting the AE's capabilities.
• Human-in-the-loop control: Ensure that AEs retain full editorial control over AI-generated content and actions, fostering trust and accountability.
• Real-time data freshness: Integrate with CRM and other data sources to ensure that the notebook always displays the most up-to-date information.
• Process enforcement through automation: Embed critical sales processes and guardrails directly into the notebook, guiding AEs and ensuring compliance.
• Continuous learning and feedback: Implement mechanisms for AEs to provide feedback on AI suggestions and notebook functionality, driving continuous improvement.
• Reduced context switching: Minimize the need for AEs to jump between multiple tools, allowing them to focus on high-value selling activities.

Permissioning & Compliance

• Row-level security: Attio API tokens scoped per rep ensure they can only sync deals they own.
• Notion guest controls: External advisors get view-only links with expiring tokens.
• Access logging: Every AI invocation or send action writes a record (deal ID, user, fields touched) to BigQuery + LogSnag, satisfying SOC 2 evidence requirements.
• Redaction filters: PII never leaves the workspace. Before text hits OpenAI, we run deterministic masking on names/emails and replace them with placeholders rehydrated in the final draft.

Compliance sits in the same dashboard as RevOps, so they can audit without a manual export.

AI Prompt Strategy

Prompts are deterministic checklists, not creative writing briefs.

• Call summaries: “List the top 3 business outcomes mentioned, blockers, decision criteria, and named stakeholders. If missing, output UNKNOWN so the rep fills it in.”
• Risk assessment: Model scores deal health (Champion strength, timing, legal blockers) and proposes mitigation tasks.
• Mutual action plan generator: Pulls stage-based tasks from a library and tailors them with deal metadata (region, product line, security posture).
• Coaching tips: When the deal stalls for >14 days, AI surfaces similar deals from the past year and what unblocked them.

Reps can tweak prompts by toggling checkboxes (“Include pricing context,” “Focus on security”). All prompt versions live in Git with code review.

Automation Moments

• After every call: Gong transcript hits Attio → triggers notebook refresh → AI drafts recap → rep edits → Resend email sent within 15 minutes.
• Stage change: Notebook posts an update to #deal-desk with the AI risk summary plus requests for legal/finance input.
• Executive briefings: Before QBRs, the system generates a one-page PDF combining ARR trajectory, mutual plan progress, and AI commentary.
• Alerting: If a notebook goes 5 business days without updates while still in Commit pipeline, the owner and their manager get pinged.

Automation freed reps from admin work without removing their judgement.

Adoption & Change Management

1. Pilot with top performers: Two senior AEs stress-tested the earliest template and gave brutally honest feedback.
2. Run “tab funerals”: We held a workshop where reps listed every tab they open in a typical day, then mapped which ones the notebook replaces. Seeing redundant tools vanish created buy-in.
3. Shadowing: New hires shadow a veteran using the desk for two deals. They learn how to trust AI suggestions while editing aggressively.
4. Weekly desk retro: 15-minute cadence where reps share wins (“AI found a blocker I missed”) and gaps (“Need procurement checklist for APAC banks”).
5. Enablement content: Loom walkthroughs for every button plus an FAQ on AI behaviors (“What if the summary misses nuance?”).

Metrics & Telemetry

• Reduced AE onboarding time: Onboarding time for new Account Executives (AEs) has been reduced by 30%, from 8 weeks to 5.5 weeks.
• Increased call note completeness: Call notes are 55% more complete, thanks to structured templates and AI nudges.
• High daily desk usage: 94% of the sales team logs activity in the notebook at least once per business day.
• Improved follow-up SLA: 92% of call recaps are now sent within 30 minutes, a significant improvement from 48% previously.
• Full audit coverage: 100% of AI-generated emails have a corresponding log entry with the human approver, ensuring compliance and accountability.

We track these metrics in Metabase dashboards embedded inside the RevOps home page and review them during weekly pipeline meetings.

What stuck with me

• AI doesn’t replace judgement; everything stays editable before sending, and reps get credit for the edits.
• Action-level logging keeps compliance happy and gives leaders visibility into deal momentum.
• Templates beat blank canvases—people adopt tools that remove work, not ones that add “write a narrative” homework.

What I'm building next

I'll be sharing this desk as a clonable template. Next on the roadmap: native Salesforce integration, multilingual recap drafts, and automated executive briefings triggered by ARR thresholds. Want it? leave me a ping.

Want me to help you replicate this module? Drop me a note and we’ll build it together.