We replaced our sales team with 20 AI agents—here’s what happened | Jason Lemkin (SaaStr)
Jason Lemkin, founder & CEO of Sastr, reveals how AI transformed his sales team from 10 humans to 1.2 humans and 20 AI agents, maintaining performance. He details the future of GTM, AI's role in efficiency, and how professionals can adapt to thrive in this new landscape.
Deep Dive Analysis
16 Topic Outline
SaaStr's Shift to AI Agents for Sales
SaaStr's AI Agent Operations and Performance
The Changing Go-to-Market Landscape in the AI Era
Future of Sales Roles: SDRs, BDRs, and AEs
How to Become Hyper-Employable in AI Sales
Why Most Companies Shouldn't Build Their Own AI Tools
Specific AI Agents and Their Applications at SaaStr
Challenges and Learnings in AI Agent Deployment
Crafting Effective AI-Generated Sales Emails
When Human Salespeople Still Outperform AI
Critical Role of Human Oversight in AI Operations
Advice for Salespeople and Founders in the AI Era
Enduring and Evolving Aspects of Sales in AI Age
AI Creates More Work, Not Less
The Incognito Mode Test for AI Opportunities
AI's Impact on Jobs and Backfilling Roles
4 Key Concepts
AI Agents
Software entities that perform tasks autonomously, often interacting with users or other systems. At SaaStr, these agents handle tasks like outbound email campaigns, inbound lead qualification, and customer support, operating 24/7 without human intervention for routine tasks.
Forward Deployed Engineer (FDE)
A role focused on ensuring the successful deployment, training, and ongoing optimization of AI software for a customer. This person works hands-on with the customer to integrate the AI, ingest data, and iterate on its performance, ensuring it works effectively from day one.
Orchestration (of AI Agents)
The process of managing, monitoring, and coordinating multiple AI agents to ensure they work effectively together, avoid conflicts, and achieve desired business outcomes. This involves segmenting the base, reviewing outputs, correcting errors, and continuously refining their operations.
Incognito Mode Test
A method for founders and leaders to objectively evaluate their company's customer experience by interacting with their own product and support systems as a new, anonymous user. This helps identify pain points and opportunities for AI implementation, especially in areas like support or sales outreach.
11 Questions Answered
SaaStr replaced a team of about 10 full-time SDRs and AEs with 1.2 human staff (one AE and a part-time Chief AI Officer) managing 20 AI agents, achieving similar business performance with significantly higher efficiency and scalability.
The fundamental 'plays' (outbound, webinars, events) still work, but traditional 'playbooks' are broken due to decelerated growth and changing ROI. AI is driving ruthless efficiency for low-end companies and enabling hyper-growth companies to service massive inbound demand they couldn't handle with humans alone.
Email-based SDRs and human inbound lead qualifiers (BDRs) are expected to be mostly extinct within a year, as AI can perform these tasks more efficiently. While 70% of AE jobs might be safe next year, this could decline to 40-50% as AI agents become capable of closing deals, especially for products with less negotiation.
Individuals should pick an agentic AI tool, deploy it themselves, and personally handle the training, data ingestion, and iteration. By becoming proficient in orchestrating and managing AI agents, they will be highly sought after by companies needing AI-savvy GTM leaders.
No, most companies should not build their own AI GTM tools. The pace of innovation is too fast, and maintaining internal solutions would be costly and quickly become obsolete. It's more efficient to buy leading vendor products and focus on effective deployment and training.
Effective AI emails are created by training the agent on the company's best-performing human-written email copy. The AI then uses this as a template, iterates through A/B testing, and personalizes emails by pulling data from CRM and website visitor tracking, resulting in 'pretty good' emails that add value.
Humans excel in high-dollar value enterprise sales requiring bespoke, deeply researched outreach, complex negotiations, and in-person interactions. For these scenarios, the ROI justifies the human effort, and AI tools can serve to augment research rather than replace the core sales function.
Human oversight is critical. A dedicated 'Chief Orchestration Officer' or similar role is needed to spend significant time (e.g., 10-15 hours/week for 20 agents) reviewing outputs, correcting errors, segmenting the customer base for different agents, and continuously iterating on their performance. Agents operate autonomously but require constant management.
AI is creating more work, not less. While it automates routine tasks, the sheer productivity of AI agents generates a massive volume of output and leads that humans must then manage, analyze, and follow up on. Professionals need to keep up with the increased pace and embrace new responsibilities.
The Incognito Mode Test involves a leader anonymously interacting with their own company's website, support, and sales processes as a fresh customer. This helps uncover pain points, inefficiencies, and areas where AI agents could significantly improve the customer experience, motivating leaders to implement AI solutions.
AI is displacing mid-pack and mediocre performers, especially in entry-level SDR and BDR roles. However, it's not primarily causing mass layoffs but rather shifting hiring priorities, leading to fewer backfills with humans and a greater demand for highly productive, AI-savvy professionals who can manage and leverage these tools.
17 Actionable Insights
1. Master AI Agent Deployment
To become hyper-employable, pick an AI agentic tool, a leading vendor, and personally commit to deploying and training it. This involves ingesting data, iterating, and spending hours correcting mistakes daily for about 30 days to achieve production-ready results.
2. Prioritize Vendor Support
When selecting AI agent vendors, prioritize those that offer strong ‘forward deployed engineer’ (FDE) or solution architect support. The best software is useless without hands-on help to train and onboard it effectively, ensuring it works on day one.
3. Buy, Don’t Build GTM AI
For go-to-market AI tools, buy from specialized vendors rather than building solutions in-house, unless you are a company like Vercel with dedicated, expert engineers. The rapid pace of AI innovation means internally built tools can quickly become obsolete.
4. Embrace AI as a Junior Professional
As a junior sales or GTM professional, proactively embrace and master the AI tools your organization uses. Don’t fight the increased transparency or productivity demands; leaning into these changes will make you more valuable and help you leap ahead of peers.
5. Leaders: Personal AI Deployment
If you are a leader or VP concerned about obsolescence, personally engage in deploying and training an AI agent. This hands-on experience (from ingestion to orchestration) is crucial for understanding the process and leading your team effectively.
6. Allocate Oversight for AI Agents
Recognize that orchestrating AI agents requires significant human oversight, even after initial training. Expect to spend 10-15 hours per week reviewing outputs and making iterations, as agents work continuously and require constant management.
7. Segment Customer Base for Agents
When running multiple AI agents, segment your customer base carefully and assign specific agents to different segments (e.g., inbound, outbound, lapsed customers). This prevents conflicts and optimizes the effectiveness of each agent’s outreach.
8. Train AI with Best Human Content
Train your AI agents using the best email copy, scripts, or content from your top-performing sales or marketing professionals. This provides a high-quality template for the AI to iterate and A/B test from, leading to more effective communications.
9. Focus on Value in AI Communications
When using AI for customer communications, focus on delivering value and ensuring an instant response. People generally don’t care if an email is from an AI, as long as it’s helpful and timely.
10. Start AI Deployment in Support
A practical starting point for deploying AI agents is in customer support. Implementing AI for 24/7 support can significantly improve response times and handle a large volume of inquiries, which most apps struggle with manually.
11. Conduct Incognito Customer Journey Audit
Regularly (ideally quarterly, at least annually) conduct an incognito audit of your company’s customer journey. Use a fresh email to test support, sales contact, newsletter sign-ups, and product usage to identify pain points that AI agents could address.
12. Embrace Higher Productivity Expectations
Sales representatives must adapt to geometrically, if not exponentially, higher productivity expectations with AI assistance. Embrace these tools to become more efficient, as fighting them will expose inefficiencies and put your role at risk.
13. Develop Deep Product Knowledge
Beyond ‘people skills,’ sales professionals must develop deep product knowledge and technical understanding. Relying solely on being a ‘people person’ is insufficient in an AI-driven sales environment where agents can also effectively interact with customers.
14. Deliver ROI Before E-Signature
In the age of AI, customers expect to see tangible return on investment (ROI) from your product before they commit to a purchase. Focus on demonstrating this value upfront, even before the contract is e-signed.
15. Founders: Evolve Current Startup with AI
For founders with existing happy customers, the best startup is often the one you’re already working at. Instead of quitting to start a new AI venture, focus on integrating AI to transform and accelerate your current business, leveraging your existing customer base.
16. Be Transparent About AI’s Impact
Be honest and positive about AI’s impact on jobs within your organization. Explain that AI will make the best people more productive and that embracing this future is necessary, rather than hiding potential changes or job shifts.
17. AI Agents for Any Scale
Don’t assume AI agents are only for large enterprises. Even companies with a few hundred customers and some website traffic can benefit significantly from AI agents, as they can handle more interactions than humans are willing or able to.
7 Key Quotes
AI is replacing the jobs people don't want to do today. And it is displacing the mid-pack and the mediocre.
Jason Lemkin
The classic SDR junior kid that is hired out of college to send emails... We don't need them. Folks that qualify leads coming in... They should be extinct next year.
Jason Lemkin
If you can go do this, pick any tool... and get it live into production, you're hyper employable.
Jason Lemkin
The net productivity is about the same. It's not better. It's not worse. But it's so much more efficient. And it scales. Because software scales.
Jason Lemkin
People were, were, were, were, were, were in an age where people don't really care as long as the email adds value. And they know they're going to get an instant response.
Jason Lemkin
The best humans, it is true that they will get superpowers from AI, but I'm not sure the rest will.
Jason Lemkin
If you're not, if you don't feel what I feel, here's my advice over the holidays... go to your app and do everything with a fresh Gmail address, try support. See how your support is. Try to contact sales... If you do this quietly, your heart, you're going to cry about some of the things you've seen.
Jason Lemkin
3 Protocols
Protocol for Becoming Hyper-Employable in AI Sales
Jason Lemkin- Pick one agentic AI tool to solve a painful or acute problem (e.g., support, SDR, inbound qualification).
- Select a leading vendor that offers good support and assistance (e.g., a Forward Deployed Engineer).
- Personally deploy the tool: train the agent, ingest your data (website URL, wiki, training docs, prospectus), and perform iterations.
- Spend 1-2 hours daily correcting mistakes and 'hallucinations' in the agent's outputs for about 30 days.
- Once proficient with one agent, apply the learned skills to deploy additional agents, which will become progressively easier.
Protocol for Training AI Agents for Sales
Jason Lemkin- Gather the best email copy from your top-performing salesperson or marketer.
- Upload this text to the AI agent as a template for its communication.
- Provide the agent with data sources (e.g., Salesforce, website visitor tracking) for personalization.
- Allow the agent to A/B test and create variants of the best emails.
- Continuously review and correct the agent's outputs daily to refine its performance and reduce errors.
Incognito Mode Test for Identifying AI Opportunities
Jason Lemkin- Open your web browser in incognito mode.
- Use a fresh, anonymous email address (e.g., a new Gmail account).
- Interact with your company's website as a new customer: try support, contact sales, sign up for the newsletter, and test your product's onboarding.
- Identify the most frustrating or inefficient parts of the customer journey where current human processes are failing or slow.
- Select an AI agent to address the most critical pain point discovered during the test.