Key takeaways
- What Perplexity Computer Actually Does.
- Why This is Different From Previous "AI Worker" Claims and what to do about it.
- Where Computer Falls Short (And This Matters).
- What This Means for Agencies and Consulting Firms.
- The Honest Assessment: Who Becomes Redundant, apply this before building anything.
I was sitting in a client meeting last week when our project manager asked the question I have heard dozens of times: “So where does this AI thing actually fit into our workflow?” I showed them a video of Perplexity Computer running. Thirty seconds in, they stopped me and said: “Wait, that is not a chatbot. That is a worker.”
They were right. On 25 February, Perplexity launched Computer, and I think it is the first credible digital employee product I have seen in five years of watching the AI space.
What Perplexity Computer Actually Does
Computer is not a chatbot with more features. It is an orchestration layer that takes complex projects and breaks them into parallel sub-tasks, each handled by the most appropriate AI model for the job. You give it a goal, “research and analyse Q1 market trends, generate competitive pricing analysis, and create a summary report” and it spins up multiple agents running Claude Opus 4.6 (for reasoning and orchestration), Gemini (for research-heavy tasks), Grok (for speed), Nano Banana (for image generation), and Veo 3.1 (for video synthesis). These run in parallel, not sequentially. It coordinates across 400+ app integrations. And it can run for hours or months on complex projects, not just one conversational turn.
The pricing is $200 per month. That is less than a single day of contractor labour in most developed markets.
Practical tip: The real value is not in replacing conversations. It is in replacing project execution for work that follows a clear playbook, market research, competitive analysis, data aggregation, report generation, first-draft content creation, workflow automation. These are the tasks that currently consume junior consultant time and generate the thinnest margins.
Why This is Different From Previous “AI Worker” Claims
I have been sceptical of the “digital employee” narrative because most products claiming this title are chatbots with marketing spin. They cannot actually manage their own time, context, or priorities. Perplexity Computer solves this differently. It treats the project as a persistent entity. It maintains context across sub-agents. It handles hand-offs between models. It can wait for external systems to respond. It does not forget what it was doing because the user closed the browser.
In one of my 120+ consulting projects, we had a team member whose role was essentially: aggregate data from 12 different sources, clean it, analyse it, and create a weekly report. That role paid $45,000 a year. A Computer subscription costs $2,400 annually, and handles that work faster and more accurately than the human ever did.
Now, the crucial part: this does not mean we fired that person. We moved them to client relationship management and strategy work. But that is the trajectory here. For roles that are predominantly execution without judgment, the floor has just moved.
Where Computer Falls Short (And This Matters)
Computer is not a general replacement for experienced knowledge workers. It struggles with three categories of work where humans still dominate:
First: judgment calls that require institutional knowledge. If you ask Computer to “restructure our sales process,” it will generate a generic framework. It will not understand your specific client dynamics, your team’s constraints, or your market position. That requires someone who has lived in your business.
Second: client relationships. Computer can draft an email. It cannot navigate a sensitive negotiation. It cannot read the room. It cannot maintain the trust and rapport that keeps a client coming back through difficult periods.
Third: novel problems that require experience to even recognise as problems. Computer will execute the brief you give it. It will not flag the thing you did not think to ask about because it has never seen your industry before.
These are the reasons that experienced consultants will not become obsolete. But entry-level roles in research, analysis, and report generation? Those are in direct competition with Computer now.
What This Means for Agencies and Consulting Firms
If you run an agency, Computer changes your cost structure immediately. Your junior team members are now competing with a $200-per-month service that does not get tired, does not miss details, and does not need training. You have two rational responses.
Response One: price your services on the value of judgment and relationships, not the cost of execution. If your advantage is “we have good junior researchers,” that advantage is gone. If your advantage is “we understand your industry deeply and we have relationships with key players,” that becomes more valuable, not less.
Response Two: reshape your delivery model. Instead of a team of two senior consultants and three juniors, you have one senior consultant, one relationship manager, and you run execution through Computer. Your margins go down on the project, but your billable rate per employee goes up dramatically.
This is already happening quietly in some firms. I know of three consulting shops that have started using similar tools to automate research phases, and they have reduced project timelines by 40% while maintaining fee levels. That is not sustainable, margins will compress eventually, but the window to restructure is open now.
The Honest Assessment: Who Becomes Redundant
Let me be direct about the timeline. Anyone in a role that is primarily execution without judgment, junior researcher, report generator, data analyst working from a playbook, content aggregator, will find it very difficult to justify their salary against a $200-per-month service within 18 months. I do not say this to be cruel. I say it because recognising the shift early is how you reposition yourself.
The junior consultants who will still have careers in 2027 are those who reposition toward judgment and relationship work. Those who try to compete with Computer on execution speed will lose.
For experienced professionals: Computer actually increases your value if you position correctly. You become the person who tells Computer what to do, reviews the work critically, and interfaces with the client. That is a higher-value role. But it requires you to let go of some of the execution work you may currently do.
The Real Story: Price Compression
The headline is “19 AI models in one platform.” The real story is the price point. $200 per month is the critical threshold. Below this price, the business case for automation becomes undeniable for any repetitive, research-heavy, or execution-based work. Above it, there is still room for human negotiation. At exactly this price, Computer is cheaper than a single contractor day.
This is the third wave of AI disruption I have tracked. The first wave was “AI can do things faster” (copilots, code generation). Most businesses largely ignored this because speed was not their bottleneck. The second wave was “AI can do things cheaper” (ChatGPT API, Claude API, open-source models). Businesses started paying attention, but it was still siloed to specific teams. This third wave is “AI can do jobs cheaper than hiring for that job,” and it is the one that forces structural change.
If your business model depends on margin extraction from junior-level execution work, you need to respond now. Not in six months. Now.
What I Am Watching Next
Three things will determine whether Computer becomes genuinely transformative or another impressive demo:
First: reliability at scale. Does it maintain consistency across 100-hour projects? Does it degrade gracefully when external systems are slow? These are engineering questions, not AI questions, and most startups are weak here.
Second: integration depth. 400+ integrations is good marketing. But does it integrate deeply with the actual systems your business runs on? Or is it mostly a thin layer that requires manual hand-offs? Integration depth is where most AI workflow tools fail in practice.
Third: cost per output, not per month. As usage scales, will the actual cost of running a project through Computer increase in ways the $200 price hides? Most AI workflow pricing has these hidden complexity costs.
If Perplexity solves these three problems, Computer is genuinely transformative. If they do not, it remains a powerful tool for specific use cases, but not a wholesale replacement for junior consulting roles.
My current assessment: I give it 70% probability of being transformative within two years.
Tagged: AI Strategy, AI Tools
I build automations, AI steps and custom software for businesses with 5 to 200 staff. 120+ projects across 15+ industries, from Middlesbrough.



