Since the release of GPT-3 in June 2020, Large Language Models (LLMs) have expanded and evolved at an impressive rate. As their capabilities have grown, they’ve captured our imaginations with their novel abilities and ease of access.
It is now clear that AI represents the most significant shift in the technology landscape since the inception of the internet. And CRM, a $58 billion category and the engine of businesses everywhere, is now having a do-or-die moment. Simply adding LLM features to legacy architectures isn’t enough.
This moment echoes a similar shift back in 1999, when Salesforce took CRM into the internet era, building it from the ground up for the cloud. Today, the category is facing another paradigm shift, but the stakes are even higher.
Now, with AI, CRM is finally about to undergo a massive transformation. At Attio, we've been building for the next era of CRM since day one.
We're going to usher CRM into the AI era, and it will represent a monumental shift for the category. Today, we'll share our vision for the next generation of CRM.
The rise of LLMs
The implications of AI for CRM are obviously profound. To understand why, we need to take a step back and look at the technology that's making this all possible: Large Language Models.
LLMs represent a monumental shift for software. They’re the foundation of modern AI, letting machines understand and generate human language in ways never before possible. They have two key capabilities:
Converting unstructured data into structured data. LLMs can take all the unstructured data we generate (like emails, messages, videos and phone calls) and turn it into structured data that’s easy to work with, like categories, summaries or action items. In other words, they can take the way humans naturally communicate and make it useful for automation, insights, and reporting.
Reversing structured data into unstructured output. Conversely, LLMs can also take structured data and turn it into natural language. They can use structured data to respond to questions, summarize complex data, records or even draft emails and messages.
Early attempts to integrate AI so far have largely focused on that second part: taking structured data and turning it into natural language. We’ve seen a lot of chatbots, copilots, and short injections of AI, like providing assistance in drafting or summarizing information. These features are useful and are already bringing meaningful productivity increases, but they’re the first step in bigger things to come.
The seismic shifts for CRM
The next big impact from AI will come from being able to convert unstructured data to structured data. For CRM, this means a platform where LLMs are natively embedded into the CRM’s core architecture and functionality, instead of being bolted on. This will create two seismic shifts for the category:
First, CRMs will gain the ability to understand and process vast amounts of unstructured data, in ways that were previously thought to be impossible.
Think about all the valuable information that's locked away in emails, video calls, phone calls and messages. Right now, this unstructured data is reluctantly and sporadically entered by humans, in bits and pieces. But with the right, AI-native data model in place, all of those insights can be automatically captured, structured, and put to work.
However, context won’t just be gleaned from language (e.g., call transcripts). Multi-modal AI models will process audio and video to understand not just words, but also tone, emotions, and the nuances of human interactions. If a client sounds excited or frustrated during a call, the AI will pick up on that. This deeper understanding will give your CRM human-like context for decision-making, as if you had your best teammate by your side for every interaction, catching all the subtle cues and providing valuable insights.
Second, your CRM will gain “agency.” Agency is the ability of an LLM to use complex reasoning capabilities to autonomously work towards a goal by breaking it down into smaller steps, deciding what actions to take at each step, and then using a set of predefined tools to execute those actions.
You’ll have multiple agents that perform complex tasks and workflows on your behalf, making intelligent choices along the way. Working together, they’ll be able to automate even the most complex business processes, including ones that require human judgment and nuance.
These two capabilities, when combined, will lead to a real paradigm shift for CRM. Go-to-market processes that used to require build-outs and complex configurations will become much simpler to implement. Tasks that previously required significant manual effort or complex automation will now be handled autonomously and intelligently by the CRM itself, but always under your direction and control. You set the goals and parameters, and the CRM executes with unprecedented efficiency and intelligence.
Our vision
For years now, CRM has been on the cusp of revolution. The traditional paradigm reached its limits a long time ago. It was never designed to handle the exponential growth of data we've seen in the past decades.
At Attio, we recognized this impending shift. Since day one, we've been deliberately building for a paradigm change, intentionally designing a powerful, flexible product with strong fundamentals, ready for the next evolution of CRM.
With this foundation in place, Attio is now going to usher CRM into the AI era, and it will represent a monumental shift for the category.
Attio will be able to completely understand structured and unstructured data and act on it, eliminating day-to-day manual data entry, automating human tasks that were previously thought impossible to automate, and becoming an intelligent partner in every aspect of going to market.
Here’s how we’re doing it:
A system of record
Building CRM for the AI era requires more than just incorporating LLMs into existing systems. It demands a fundamental reimagining of the CRM data model – one built from the ground up to work natively with AI.
This is where Attio Objects comes in. We built this powerful data model to be the modern system of record for businesses. It has custom objects, so it can match any business or data model, and it can handle massive workloads with millions of records with 50 milliseconds or less of latency.
But what really sets Attio Objects apart is how its fundamental design lends itself so well to AI models. We built it to be flexible and structured, with precise data types, rather than treating everything as simple strings. Each property in our data model holds semantic context about the information it represents.
This approach, which was crucial for us in creating a powerful system of record, is also incredibly important for AI integration. When an LLM interacts with Attio, it's not just seeing a collection of data – it's interfacing with a structured representation of your business.
When describing your data structure to an LLM, Attio Objects doesn't present just a random collection of information. Instead, it offers a rich semantic graph of your business. This allows the AI to truly understand the context of each piece of information. For instance, the LLM can distinguish that a particular field represents a currency value in USD, or recognize the difference between an email address and a phone number.
Fundamentally, Attio’s data model doesn’t just give a powerful way of storing information, but also a sophisticated method of representing it. This representation is invaluable to LLMs when they're reasoning, summarizing, or automating tasks, as they can understand that a specific field has a particular meaning and adjust its operations accordingly.
A system of context
The CRM should be the all-knowing, central source of truth for your business. But the reality is that no matter how many integrations you have, it still needs context to function (which is almost always manual). Think how many sales reps have data entry stipulated in their performance goals, or even in their contracts. And without that data, your business can’t move.
In the near future, Attio will automatically ingest and understand all of your data – both structured, like we do now, and unstructured data – the details and aspects of every video call, meeting, email, document, or even data from the web. As mentioned earlier, this goes beyond language. Your CRM will know not just the meaning of words being spoken, but also the tone and body language of the customer.
Your CRM will autonomously and consistently extract insights, populate attributes, and trigger workflows based on the content of emails, video calls, and other customer interactions. As a result, Attio will be able to understand and process data in ways that were previously thought impossible.
For example, imagine a feature that can automatically derive the value of a deal based on what was said during a sales call. By ingesting the unstructured data of the call in real time, understanding the context, and cross-referencing with external data sources and your CRM data, Attio could intelligently infer the deal size – without the sales rep ever having to manually enter that information in an attribute.
For the first time, manual data entry will no longer be a barrier to having a complete, up-to-date view of your customer relationships.
A system of action
Once you have complete context, you can then use Attio’s intelligent agents to drive every GTM process, even ones that were thought to be impossible to automate because they require human judgment.
Attio will be the place where you architect your entire GTM strategy. Our vision is that built-in AI agents will help you run your business so you can focus on high-value, high-impact work. They will take the manual work out of your daily business workflows and processes, even if they involve human judgment or input.
Early attempts at standalone agents for businesses are already available, offering anthropomorphic helpers like “Tom” and “Eva,” ready to assist you with your business. While these may prove to be useful tools, their long-term potential is capped by their specificity and the shape of their product.
We believe that the future of AI in CRM does not mean having an army of virtual assistants at your disposal. These standalone helpers, while novel, are ultimately limited by their separation from your core business data and processes.
In Attio, intelligent agents will not be personified, and they will be intricately woven into the platform. They’ll automate even the most complex tasks and processes, and continuously learn and improve.
They’ll be seamlessly incorporated into the CRM, working in the background to help your business run. These will come by default in the product, configurable under the hood, and with the ability to create as many as you want. You’ll have fine-grained control over how they operate.
Embedded intelligence
The goal is to make you more productive and powerful, but in a way that feels natural and effortless. You won't be constantly reminded that you're interacting with AI – you'll simply find yourself accomplishing more, with less effort.
Attio’s AI capabilities will extend beyond action and data ingestion: they’ll be embedded into nearly every interaction a user has with the platform. This will create a completely new type of user experience.
Attio will provide intelligent suggestions and even guidance based on your interactions as you work. When you’re on a call with a prospect, Attio will take notes, pick up on verbal cues, and then suggest a specific case study to mention based on something your prospect just said.
Our platform will be an intelligent partner at every step, making dynamic, context-based decisions based on the real-time ingestion of all your data, and even adjusting the interface based on how you work. It will continuously learn and improve over time, adapting to your unique business needs and processes.
You’ll also be able to interact with Attio using natural language commands and questions for virtually every aspect of running your business, from building sales pipelines to creating new workflows.
The goal is to make you more productive and powerful, but in a way that feels natural and effortless. You won't be constantly reminded that you're interacting with AI – you'll simply find yourself accomplishing more, with less effort.
This is the future of CRM user experience – an evolution that amplifies your capabilities while fading into the background. It's AI that empowers you, without overshadowing you.
The next generation of CRM
The arrival of LLMs has sparked a lot of discussion about how CRM will evolve. The most common refrain is that CRM is going to go from being a system of record to a system of action.
But this isn’t quite right. To take action, you need context. And to have context, you need data. The truth is that the next generation of CRM will need to be all three of these things.
At Attio, we're building a CRM that doesn't just ingest data and act, but also understands, learns and adapts. It's a closed-loop system where every action will feed back into the system, informing future decisions and constantly refining its understanding of your business.
Just as Salesforce ushered in the CRM for the cloud era, Attio will do so for the AI era.
For the first time ever, CRM will finally be able to deliver on its promise to not just deeply understand your entire business, but to actively help you run it – and get better at doing so over time.
