Ex-McKinsey expert Richard Karlsson on how AI can change the consulting industry
Backed by investors including Octopus Ventures and Yanno Capital, Grasp serves almost 200 customers across 30 countries, including all Big Four consulting firms. CEO Richard Karlsson explains how the Swedish AI start-up is automating investment banking and management consulting work using multi-agent systems.
It was another late night in the office, hunched behind a monitor, manually tagging and segmenting companies. Trawling through an endless list of company websites or public filings. Building PowerPoints slide by slide. This was the unspoken assumption that if you work in consulting or finance, this is just what you do. You stay late, you pull all-nighters, you chew through 200-page reports because that is the job. The grind is the price of admission to a prestigious industry.
The daily instructions were often straightforward: to find companies in a specific value chain, tag them correctly, and segment them based on products and positioning. In practice, this meant a team of four people spending a month going through company websites one by one, manually mapping and categorising them so that we could build the analysis the client was paying for.
A market that represents $1.4 trillion in annual spending, remained almost exclusively human-driven, usually performed by teams of junior employees, whose work is 90% mechanical and repetitive. The great irony of that is we were doing the kind of work that is obviously ripe for automation, like large-scale company mapping, market landscaping, and benchmark analysis. But no one was seeing just how much of that work could be done differently.
Gap in the market
In 2019, my colleague Johan and I were experimenting with what at the time was the cutting edge of AI, but by today’s standards, rudimentary wouldn’t even cover it. Practically ancient would be more apt. We wanted to gauge the power of AI as a tool for harvesting and analysing data, so we decided to model it on what we knew best, our jobs. Over one weekend, myself, Johan and former Ericsson AI engineer Simon, reproduced much of what four consultants had done over four weeks. It wasn’t perfect, but it was the first moment that it became obvious to me that a huge part of consulting work can be automated.
As we kept experimenting, we saw models improving almost week by week. Yet, while colleagues absolutely recognised the workload, no one really seemed to connect the dots. When we launched Grasp AI, I don’t think anyone at McKinsey mentioned AI to me.
The problem wasn’t that AI didn’t exist, but that it didn’t fit the traditional model of consulting, a model that was resistant to change. Consulting is built on the mantra that time is money, with large teams billing a lot of hours, so there isn’t a strong internal push to ask whether a smaller team with an AI system could deliver the same outcome.
Research teams at large advisory firms are a key example of this. Across the industry, these kinds of organisations employ thousands of people whose job is essentially to search, extract and lightly structure information so others can use it. But, having thousands of people doing work that can be so simply automated is untenable.
In a typical high-end consulting engagement, most of the effort still goes into assembling a 100-slide PowerPoint, full of market data, competitor overviews, company financials, and competitive dynamics. We’ll look back in ten years and wonder why we ever used teams of four to six people to do that manually, when we can have all the relevant data, a map of the competitors, and visualisations of key insights in seconds at the press of a button.
If AI is going to remove much of the cost base, margins will initially go up, but over time, as competition increases, prices will come down. Threatening as this sounds for the consulting industry, Jevons’ paradox suggests that when the price of something goes down, demand rises, so total consumption actually increases. As we saw with cloud computing, cheaper data storage meant companies simply stored more data. If AI radically reduces the price of data and insight for decision-making, the number of decisions taken with that level of support can explode.
Important investment
Today, a lot of companies simply never enter the market for high-end advisory because it is too expensive. My family runs a company, so they make strategic decisions about markets, growth, and expansion every single day. But they would never spend money on consultancy, even though they may well have questions a McKinsey team could help answer. If an AI system can provide insights at a low cost, even companies with just a few million dollars in revenue will be able to base their decisions on data rather than instinct.
The big consultancies will see the price per project fall, but the number of potential projects can increase. So rather than being the person who manually produces charts or pulls information from databases, consulting work will shift towards becoming a sparring partner, understanding what the client is really trying to achieve and challenging assumptions. Advisors will work on more projects in parallel, with the research, data gathering and visualisation largely automated.
When I think back to those nights at McKinsey, poring over market reports and websites, the most surprising thing in retrospect is not the inefficiency. It is the absence of a conversation about alternatives.
AI is unbundling the consulting value chain, automating research divisions, compressing deck-building cycles from months to minutes, and giving end clients tools that used to exist only inside the advisory firm’s firewall. The firms that thrive will be the ones that acknowledge this and rebuild around it. The firms that don’t will see value, margin, and client power shift away from them to AI-first players and, increasingly, to clients themselves.
And they will look back, not too many years from now, and realise that the most expensive decision they made was to say nothing at all.
