From Chatbots to Thought Leaders: AI's Quantum Leap in Sales
Sales professionals are one of the earliest adopting commercial user groups of AI. As of 2024, 50% of sales professionals view AI as critical to their work, and 78% believe AI tools enable them to focus on more important and strategic aspects of their jobs.
This strong positive sentiment is unsurprising. Sales is ripe for AI disruption for three, key reasons.
First, sales organizations generate an incredible amount of data (sales call recordings, email exchanges, purchase history, lead information, etc). Today, the majority of this data goes unutilized, independently or comprehensively, to drive better sales decisions and insights.
Second, many AI sales tools can deliver huge efficiency gains without requiring 100% accuracy. As an example, drafting outbound email campaigns is one of the most common, early AI use cases in sales. Tools that assist with email drafting can incorporate customer data from private and public sources, and easily keep a “human in the loop” as a stop gap until accuracy improves, with relative ease. (Importantly, many of these tools can be used in a “prosumer” capacity, without requiring broader implementation at a team or company wide level).
Third, sales professionals spend an inordinate amount of time on mundane, repetitive, non-nuanced tasks (e.g. data entry, CRM updates, meeting scheduling, and handling repetitive customer requests.) Importantly, this is the case for more senior sales professionals, such as solutions engineers and AEs, as well as more junior positions like BDRs and SDRs.
The diagram below illustrates a rough calculation for a 15 person sales team. As shown, a sales organization can easily waste ~25% of total spend on “ops” which is largely inclusive of automatable work.
Status of AI in Sales: Bots > Co-Pilots > Agents
Sales co-pilots, or “bots” as they were originally known, have been around for over a decade. These companions generally follow predefined sets of rules and conditions structured in a hierarchical, decision tree format. While accurate and consistent within specifically designed scopes, these bots are highly limited in that they can’t handle ambitious or unknown user inputs. They also can’t learn or improve based on user interactions.
Over the last few years, “bots” have evolved to co-pilots, capable of working alongside humans to handle more complex tasks. This transformation is underpinned by advancements in LLMs and GenAI, enabling co-pilots to interpret user intent, understand nuances, and maintain coherent conversations. Unlike rigid chatbots, AI copilots are highly flexible, capable of adapting to various contexts and user needs.
Today, most incumbent sales software providers have launched some form of a co-pilot that lives within their existing platform. Hubspot’s Breeze co-pilot, for example, helps sales reps write highly customized outbound emails leveraging customer data stored in Hubspot.
While co-pilots are delivering huge productivity gains to early adopters, they’re ultimately still reliant on ongoing human oversight. This fundamentally limits their ability to automate full workflows that require the approval of sequential steps. Over the last year, advancements in infrastructure and platforms has resulted in the possibility that AI agents will emerge as the next wave of AI sales applications.
Unlike co-pilots, AI sales agents can operate autonomously to complete specific goals without human intervention. They often work within frameworks, or agentic systems, where each agent is specialized in a specific task, the sum of which results in the completion of a complete goal or workflow. Importantly, unlike co-pilots, AI-agents don’t need to “live” within existing platforms, and don’t require interfaces designed for human consumption. They can be cross-platform, absorbing knowledge, data, and context from limitless sources.
Over the last 12 months, we’ve seen the rise of agentic systems going after supporting sales infrastructure products, like CRM, to specific functions of the sales process, like prospecting. Attio, as an example, is building a net new “CRM” that completely re-imagines sales data utilization from a multi-model and unstructured perspective. Others like Clay and 11x are creating agentic systems for outbound sales orgs, effectively eliminating the need for BDRs.
Contextualizing Co-Pilots vs AI Agents: Salesforce
To contextualize the difference between AI co-pilots and AI agents in sales, let’s look at Salesforce’s evolution from co-pilots to agents.
Co-Pilot: Salesforce’s Einstein
Salesforce released its first co-pilot, Einstein, in 2016. Einstein lives within Salesforce's existing UX and assists human sales reps with clearly defined tasks that require ongoing human feedback.
Sales meeting prep is a typical Einstein use case. A rep will prompt Einstein with a request such as “help me prepare for my meeting with client X”. Einstein processes this request by identifying the Salesforce record for client X and retrieving and analyzing all relevant account data (e.g. open opportunities and stages, customer service cases, transcribed call notes, etc.). After synthesizing and summarizing account information, a sales representative may request a subsequent set of actions from Einstein, such as drafting an agenda and accompanying email to send to participants ahead of the meeting. Importantly, Einstein requires continual feedback and is predominantly reactive vs proactive.
AI Agent: Salesforce’s “Agent force”
“Agenforce”, unveiled during Dreamforce 2024, takes Salesforce’s AI assistance to a new level with its suite of autonomous AI agents. These agents can operate independently to complete multi-step tasks, and interface directly with current and prospective customers.
As an example, Salesforce offers an “SDR agent” to engage with inbound leads and “set meetings with an AE”. In natural language, users can communicate the instructions around how to complete that task, such as “determine willingness to pay and escalate contracts that could be greater than $10,000”. The SDR agent engages via email with prospects, answering any pricing and product questions using documentation uploaded to Salesforce's vector database, ahead of suggesting times that work for the AE and sending the appropriate calendar invites. Ahead of the meeting, the SDR agent sends a summary of the conversion to the AE, suggesting topics to cover and potential key objections.
Unlike co-pilots, Salesforce’s AI agents are treated like human employees. They can be assigned managers who track their progress and can offer feedback and “coaching”.
Looking Forward: Incumbents VS Startups?
There is an ongoing debate about who will “win” the AI race at the application layer, in sales as well as other verticals like HR, marketing, ops, and more. While incumbents are keenly aware of the need to incorporate AI into their offerings, many are doing so in a way that is constrained to the confines of current SaaS products and user interfaces. This results in AI being layered onto existing products rather than fundamentally reimagining the functions and workflows these tools were originally designed to assist humans in completing.
This moment in time, driven by unprecedented advancements in AI, is not the first time tech incumbents have faced the decision of whether to cannibalize existing, successful product lines to fully realize the value of a new technology or platform shift. However, the headstarts awarded to legacy providers during prior tectonic shifts (e.g. on premise to cloud or desktop to mobile) are severely less advantageous this time around. Three key examples include incumbents’ (1) vast resources (2) “systems of record” and (3) deep expertise in human workflows.
Vast Resources. AI-powered tools, from code generation to marketing and design, are enabling startups to operate with unprecedented speed and efficiency. Startups today can build products 10x (if not 100x) faster, facilitating rapid prototyping, quick iterations, and faster market entry. This acceleration in development cycles enables startups to compete with established players in areas that were once resource-intensive and time-consuming. My partner Jeff is writing a book on leveraging AI in GTM you can learn more about here. (It’s worth noting that incumbents are also rapidly adopting AI tools, but not at the pace of earlier stage companies unencumbered by legacy tech, bureaucracy, more substantial privacy and security concerns, etc.)
“Systems of Record”: Incumbents have historically benefitted from being “systems of record” for customer data. In the previous era of structured, text-based databases, the complexity of migrating data elsewhere had to be accompanied by 10x+ improvements in functionality. As such, despite incredible developments in supplemental tooling for specific industries, “system of record” incumbents have remained relevant and critical (there are dozens if not hundreds of apps used in conjunction with customer data stored in Salesforce). LLMs completely shift this paradigm, unlocking the full potential of multimodal, unstructured data, and severely limiting the relevance of existing databases and record systems. (In fact, AI agents often “live” a step ahead of systems of record, completely transforming the need for traditional databases altogether).
Deep Expertise in Human Workflows. As AI systems become increasingly capable of executing complex tasks without human oversight, the current incumbent platforms that enhance human efficiency and collaboration will become irrelevant. The next generation of application tooling won’t have traditional interfaces centered around human input and control. As an example, the concept of "fields" in traditional CRM systems will become obsolete. Instead of sales representatives manually inputting data into predefined fields, AI agents will continuously gather, analyze, and act upon vast amounts of data from various sources.
While layering co-pilots and agents into existing sales software augments human labor and is clearly valuable, these productivity gains are incremental versus step-change improvements. It may take years, but incumbents that don’t completely re-imagine how sales should and could function in a genAI world, and design products for that future, will be unseated by a new wave of offerings.
Final Thoughts: How Will Sales Roles and Teams be Impacted by AI
We’re closely watching how sales teams evolve as they incorporate AI into their workflows to augment human labor. (See the below graph which shows various sales functions and the degree of automation today). A few team evolutions we’re already witnessing, or predict in the near future, include:
Sales professionals becoming “full stack”. Tools that automate the work of SDRs will be used by AEs, for example. As such, the role of AEs will evolve, not only to encapsulate the SDR function, but also likely “managerial” skills to monitor, coach, and train AI SDR agents.
The reduction/elimination of BDRs and SDRs due to AI-automated outbound will require the creation of new entry level sales training methodologies. Entry-level sales roles serve as vital training grounds for AEs today. Sales teams will need to rethink sales training as lower-level roles that previously served as valuable training grounds are automated away.
The “pre-sales” function will fundamentally change. Sales Ops may become the “command center” of outbound teams, equipping a set of task or channel specific agents to execute targeted outbound. Just as we’re seeing the emergence of “10x” founders, we expect to see the same multiplier effect for sales leaders. My partner Jeff writes more about this concept here.
If you’re building in the AI GTM space (particularly if you’re completely re-imagining how a sales organization could function leveraging AI) we’d love to meet. Please shoot a note to julia@flybridge.com.
