As companies implement AI into more and more workflows, executives need to double down on human context to get the most out of the technology, something that’s going over the head of some leaders.
“A significant amount of people don’t realise that we can have the best AI model in the world, but without the right context about what’s actually happening, you won’t get the outcomes you’re looking for,,” says Carolina de Freitas, who previously led marketing and advertising technology at gaming operator Allwyn.
Without this context layer, organisations deploy AI algorithms that generate results “without really understanding the context behind the user behaviour,” she adds.
Context-driven AI is changing how businesses operate. Instead of delivering generic messages to a broad group of customers, AI can learn customer interests and habits and use them to tailor what they see.
But how does this work in practice?
The trap of the ‘smartest model’
Without context, AI models won’t give companies the outcomes they’re looking for, says Jonathan Lakin, CEO at AI and behavioural intelligence company Intent HQ, which analyses behavioural data to help clients predict what customers will need or do next.
Today there is a limit to how Large Language Models (LLMs) — such as OpenAI’s ChatGPT and Anthropic’s Claude — process information. The maximum amount of information an AI model can hold, process and reference during a single interaction or request is called a context window.
Human context is the layer of understanding that gives AI meaning behind customer data.
Although a generative model can draft content and analyse large amounts of data, “an LLM can only hold a short context window,” adds Lakin. “The largest context windows [for commercially available models] now are about a million tokens.”
Without personal and behavioural context, AI systems often recreate human blind spots at scale. To get the most out of a model, companies need to feed it more human context and personalised and behavioural data.
“Human context is the layer of understanding that gives AI meaning behind customer data, helping it understand an individual's behaviours,” says David Porras, an AI and data analytics executive at Fifth Third Bank.
This is why business leaders are shifting away from focusing on raw computing power alone as a measurement of a model’s ability.
Intent HQ’s customers include consumer-facing enterprises, such as banks, telecommunications companies and retailers.
Clients embed Intent HQ’s technology directly into their own existing mobile or device app. Once a customer downloads or updates the app, Intent HQ’s software runs in the background, analysing daily habits, locations, and routines locally.
For example, instead of a shoe retailer sending a discount to everyone who bought shoes, Intent HQ’s technology allows it to understand exactly what type of shoes a customer is looking for based on their lifestyle and send them relevant messaging.
The missing context layer
To bridge the gap between raw data and AI that understands human context, systems require an infrastructure layer focused on what Lakin calls “deep signal intelligence.”
Translating the everyday actions of a user into usable context means processing large volumes of data, he says. Deep signal intelligence is the ability to sift through this behavioural data to figure out which specific actions reveal human intent.
"We need to be able to do that on a device, without any raw data leaking.”
Historically companies have accessed customer context and information by hoarding data in data lakes — centralised locations that store data in its raw format until needed for analysis.
But that approach is becoming increasingly difficult. Global data protection laws means customer profiling is restricted and storing behavioural data creates a prime target for cyberattacks.
“Data protection isn’t going anywhere,” says de Freitas, “and so the less customer data we hold in our data warehouses and internal systems, the better for regulatory compliance and risk management.”
Context-native companies — which use customer behaviour and habits, rather than raw computing power, as the foundation of their AI strategy — are adopting decentralised, on-device systems to store data.
We need to be able to do that on a device, without any raw data leaking.
"If you think about how Apple handles privacy,” says Lakin, “which is on-device, that's how we think about it."
Instead of transferring all of a user's raw data such as patterns, routines or app usage to a data warehouse, the data remains on the user's device.
Rather than a customer of Intent HQ giving its AI system a generic targeted message to send to its own potential customers, the backend computer system broadcasts a request.
For example, instead of tracking a user's location 24/7, the system might ask all devices: ‘Do any of you belong to a user whose evening train is delayed, and who usually orders food delivery’ — a food delivery platform might then use this to trigger a localised discount advertisement.
Software on the individual's device analyses its own data and if the context matches the criteria of the request, the phone accepts the message and displays it to the user.
How context-led AI works in practice
Traditional AI software is run on human-built rules and guardrails. Marketing automation, for example, relies on binary triggers — if a user clicks a specific link, a system automatically sends a pre-written follow-up email.
These actions take up a lot of time, says de Freitas. Legacy systems "took years to build because they were built by humans, each with their own ways of working. It can be challenging to move people away from those established approaches.”
If you have the right context, you can understand what matters to a particular group of people in a particular situation.
Rather than just tracking what a customer does, context-driven AI focuses on the motivation behind the action.
One of the most important parts of using human-context AI is knowing when not to act. Without context, companies target their entire audience with generic and sometimes inapplicable campaigns.
"If you have the right context, you can understand what matters to a particular group of people in a particular situation,” says de Freitas.
By applying a contextual AI layer for a telecommunications client, Intent HQ helped to reduce spam messaging.
"We took the number of messages they sent from 400 per customer per year to 65,” says Lakin.
“We drove and increased their revenue,” he adds, because Intent HQ could tell its customers not to target specific people.
The shift to proactive foresight is also beneficial in financial services. “Banks already possess an enormous amount of customer data,” says Porras. “But traditional systems primarily organise, report and analyse what has already happened in the past."
For example, instead of waiting for a client to log in and request a loan, banks can identify shifts in customers' financial circumstances in real time.
“Adding human context data enables AI to interpret individual behaviour, patterns, interests, intentions and the potential needs of each customer,” Porras adds. “Providing a better client experience in a more proactive, precise approach.”
Intent HQ’s IntentOne platform was recently named AI Solution of the Year (International) at the AI World Series Awards 2026. Read more here.
