AI-First Strategy: How Kaddi Is Executing

AI is no longer the hard part. The real advantage now comes from how well a company can wrap AI around the way its business actually works. In B2B sales, that means more than adding a chatbot or automating a few admin tasks. It means codifying the sales process, connecting it to the buyer journey, embedding the right content and methodology, and helping sales teams execute the right moves on real deals. Inspired by Tom Tunguz’s idea of the “AI harness”, this post outlines the seven key components of an AI-first strategy — and how we are applying them to help sales teams turn process into performance

1. Context & Memory

Our first priority is to make AI deeply contextual to each sales organisation. Kaddi is built around the idea that every company has its own way of selling: its stages, steps, exit criteria, key actions, methodology, buyer journey, content, qualification rules and coaching standards. We are executing against this by turning each customer’s sales process into a living “context database” that the AI can draw from on every deal, so the guidance reps receive is not generic sales advice — it is grounded in how their business actually wins. This aligns directly with Tunguz’s point that AI needs bespoke retrieval and a “recipe book” for how the business runs. 

2. Tools & Action

An AI-first sales platform must do more than answer questions; it must help reps take the right action at the right moment. Kaddi is executing this through practical sales tools that support deal progression: draft emails, draft slide decks, meeting prep, meeting summaries, deal health checks, MEDDPICC support, stakeholder research, content recommendations, proposal support and next-best-action guidance. The goal is to connect the AI to the real work of selling, while keeping the seller in control of sensitive buyer-facing actions. In Tunguz’s terms, the process tells the AI what should happen, while tools give the AI a safe way to help make it happen. 

3. Orchestration & Loop

We are not treating AI as a one-off chat interface. Kaddi is being designed around an execution loop: understand the deal, assess what has happened, recommend the next move, support the seller in taking action, then learn from the result. This means AI becomes embedded into the rhythm of sales execution - meeting prep, discovery, evaluation, proposal, negotiation, contract, close and post-deal review. Our assessment, execution analysis, blueprint and pilot model gives us a structured way to test this loop with real reps and real deals before scaling it across the sales team. 

4. State & Persistence

Complex B2B sales do not happen in a single session, so the AI must remember where each deal stands. Kaddi is executing against this by building persistent deal memory across transcripts, emails, notes, documents, contacts, shared files, CRM data and process status. This allows the AI to pick up the thread of a deal, understand what has already happened, identify what is missing and guide the seller from the current state rather than starting from scratch every time. For sales teams, this is essential because opportunities often involve multiple stakeholders, long cycles and many handoffs.

5. Sandbox & Compute

For AI to be trusted in sales, each customer needs a secure, controlled environment where their data, process and content remain protected. Kaddi is executing this through client-specific workspaces and controlled data handling across uploaded content, sales process assets, deal records and integrations such as Google Drive or Salesforce. The key principle is that customer knowledge stays customer-specific: their sales process, customer insights and deal data are used to support their execution, not to train or benefit other customers. This is the practical foundation for making AI useful in enterprise sales without compromising confidentiality or trust. 

6. Observability & Governance

AI-first sales execution needs visibility, not blind automation. Kaddi is executing against this by making process health, deal quality and rep execution measurable: strengths and weaknesses, content gaps, crucial steps, sales process health KPIs, deal recommendations and coaching priorities. The aim is to give sales leaders a clear view of whether the team is following the process, where deals are at risk, what content is being used, and where human coaching is required. This matches Tunguz’s point that production AI needs tracing, evals, guardrails and human oversight — especially where judgement, forecast calls and customer commitments are involved. 

7. Cost & Workflow Optimisation

Our AI-first strategy is not to automate everything. It is to be deliberate about which jobs are best done by AI, which jobs are best done by humans, and which jobs work best as a human-plus-AI workflow. Kaddi is executing this through the Human v AI Jobs to Be Done matrix, helping customers decide where AI should draft, analyse, summarise, recommend or retrieve — and where the seller or sales leader must make the call. The commercial focus is simple: use AI where it improves execution quality, saves rep time, strengthens coaching and helps lift win rates, deal size and deal volume