AI Wants My Job. Who Is the Real Tool?

For most of the modern business era, technology was something a company owned, installed, licensed, or rented.

In the 1980s and 1990s, technology behaved like machinery. Companies bought servers, databases, networking equipment, desktop software, accounting systems, and productivity tools. It was expensive, fixed, and visible. You owned it. You depreciated it. You hired employees to operate it. The relationship was clear. Management made decisions. Employees used software. Software executed instructions.

Then came the SaaS and cloud era of the 2000s and 2010s. Companies stopped buying software outright and started renting it. CRM, payroll, accounting, collaboration, marketing, analytics, data storage, and infrastructure moved from the server room to the subscription invoice.

This changed the economics of technology. Capital expenditure became operating expenditure. Software became easier to adopt, easier to scale, and easier to abandon. But the underlying relationship did not really change.

A CRM system did not decide the sales strategy. Accounting software did not negotiate with suppliers. Marketing tools did not wake up at 3am and redesign campaigns. Cloud software made companies faster, leaner, and more flexible, but it remained mostly passive.

The 2020s are different.

AI agents are the next migration. They move software from tool to participant. They do not simply wait for an employee to press a button. They can initiate tasks, coordinate workflows, retrieve information, draft responses, write code, analyse documents, contact customers, schedule meetings, trigger other systems, and increasingly act with limited supervision.

This is not just better software. It is software crossing the boundary into work.

That is the point many small and medium-sized businesses need to understand. The agentic economy is not really about whether AI can write a better email, summarise a meeting, or produce a marketing plan. It can. That argument is over.

The real question is more uncomfortable: how much control is a business willing to concede to software?

OpenAI describes its agent products as systems that can “do work for you independently,” using tools and software interfaces to complete tasks. Salesforce markets Agentforce as a way to build autonomous agents that support employees and customers around the clock. ServiceNow describes AI agents as an “autonomous workforce.” The language is not accidental. The market is no longer selling software that helps employees work. It is selling software that starts to do the work itself.

That is a profound commercial shift. Tools have bounded costs. Staff create activity. A software licence sits quietly until someone uses it. An autonomous agent can keep generating work, consuming compute, calling other models, updating systems, escalating tasks, contacting customers, and producing second-order consequences that nobody explicitly requested.

The cloud era introduced surprise infrastructure bills. The agent era may introduce surprise operational behaviour.

This is not a reason to panic. It is a reason to think clearly.

The first serious choice is whether agents are tools, employees, or operating infrastructure. Most companies will begin with the first and drift toward the third without consciously deciding to do so.

At the tool level, AI is relatively familiar. An employee asks for a draft, summary, spreadsheet, research note, customer response, or code snippet. The person remains in control. This is the restaurant model. You choose the restaurant, sit at the table, read the menu, and decide what to eat. The meal is prepared by someone else, but the experience is still yours.

The second stage is closer to delivery. You choose the cuisine and perhaps the restaurant, but the experience is mediated. You do not see the kitchen. You do not speak to the waiter. You receive the output because the backend systems coordinate the kitchen and delivery logistics. That is where many businesses are today. The employee defines the objective, while the software coordinates the workflow and execution.

The third stage is more unsettling. Your granola and Greek yogurt breakfast arrives at 6am because a system knows your calendar, blood sugar, sleep quality, weight target, and productivity goals. It is probably healthier. It saves time. It removes poor choices. But the aesthetic experience of choosing, eating, and living has been narrowed. The system has not removed choice by force. It has removed it through optimisation.

That means no more cream cheese bagels and ristretto for you my friend.

This is the real tension in the agent economy. Not good versus evil. Not progress versus nostalgia. It is efficiency versus control.

The same logic applies to work. First, you buy a car because you need to drive to the office. Then you use Uber because driving is inefficient. Eventually, no car. The meeting no longer requires you at all. The system attends, records, summarises, responds, and updates the project plan. The output may be better. The business may move faster. But a question remains: what part of the work was actually the work?

This matters because businesses are not only output machines. They are social structures. They carry judgment, taste, trust, memory, relationships, and culture. In many firms, especially smaller ones, the difference between good and average is not process efficiency. It is the founder’s instinct, the account manager’s relationship, the technician’s pride, the salesperson’s timing, or the way the company makes a customer feel when something goes wrong.

Automation can improve all of this. It can also flatten it.

A company may save money by replacing human contact with synthetic responsiveness. It may answer faster. It may follow up more consistently. It may never forget a ticket, a birthday, a renewal date, or a complaint. But if every customer interaction becomes optimised, synthetic, and emotionally hollow, the business may become more efficient while becoming less valuable.

Some customers will accept that. In low-trust, low-emotion, transactional markets, they may even prefer it. Nobody needs poetry from an airline baggage chatbot or a utility billing system. But in relationship-driven businesses, high-value sales, advisory services, hospitality, care, education, and complex problem-solving, the human layer is often not a cost centre. It is the product.

This is why the future is unlikely to be all staff or all agents. The winning businesses will be those that define the boundary properly.

The practical way to think about adoption is not as “use AI” or “do not use AI.” That is too crude. The better question is: what level of control should the agent have?

The first layer is automation without judgment. These are repetitive, operationally expensive, emotionally low-value activities: meeting notes, document filing, invoice matching, scheduling, CRM updates, internal reporting, first-pass research, knowledge retrieval, basic analytics, and administrative coordination. Most SMEs should automate these aggressively. Not because AI is fashionable, but because skilled employees should not spend their time moving information between systems.

The second layer is augmentation with accountability. This is where agents improve employee productivity without replacing decision-making. They help draft proposals, prepare sales meetings, synthesise research, monitor workflows, support customer service, accelerate software development, and produce first-pass analysis. The employee remains accountable. The agent increases throughput.

This is likely where the largest sustainable gains will emerge. A five-person team can begin to operate with the discipline of a fifteen-person team. A founder can prepare better, respond faster, and manage more complexity without hiring too early. Used properly, agents can give SMEs access to operational capacity that historically belonged only to larger firms.

The third layer is delegated interaction. This is where the agent begins to touch customers, suppliers, staff, or external stakeholders directly. The commercial value can be real. Response times fall. Follow-up improves. Processes become more consistent. But the risk rises sharply because the agent is no longer operating inside the machine room. It is representing the business.

This is where many companies will make mistakes. They will automate customer-facing work because it is easy to measure the savings and harder to measure the loss of trust. They will reduce friction and accidentally remove character. They will improve responsiveness and weaken relationships. They will sound more intelligent and feel more empty.

The fourth layer is autonomous operation. This is where agents initiate actions across systems with limited supervision: launching campaigns, changing pricing, approving workflows, escalating complaints, adjusting inventory, creating purchase orders, moving data, or triggering payments. At this level, the agent is no longer a productivity tool. It is part of the operating model.

That is where governance becomes non-negotiable.

The most important question is not what the agent can do. It is who is accountable when it does it.

If the answer is a person, the system can be managed. If the answer becomes a workflow, a model, a vendor, or “the AI,” the business has a problem. Accountability cannot be delegated to software. It can only be assigned to people.

This is the red line. A business can delegate activity. It cannot delegate responsibility.

That may sound bureaucratic. It is supposed to. Once software behaves like labour, it needs management. Once it can act across systems, it needs permissions. Once it can affect customers, it needs standards. Once it can create cost, it needs budget control. Once it can influence decisions, it needs governance.

This is where the agent economy becomes less glamorous and more important. The hard work is not writing clever prompts. It is designing the operating model.

Who approves external communication? Who reviews sensitive outputs? Who monitors cost? Who defines escalation rules? Who can override the agent? Who audits what happened? Who carries the blame when it goes wrong?

These questions are not obstacles to adoption. They are the conditions for using agents responsibly at scale.

The danger is not that AI agents become evil. That framing is childish. The nearer risk is organisational drift. A company slowly loses visibility into why decisions were made, why costs increased, why customers were contacted, why workflows escalated, or why the business started behaving differently.

Managers may begin supervising systems whose internal reasoning they do not fully understand. Employees may defer to outputs because challenging them takes longer. Customers may interact with a company that appears responsive but no longer feels present. Over time, the firm becomes more automated, but less governed.

The companies that get this right will not be the ones with the most automation. They will be the ones with the clearest command structure.

They will know where agents are allowed to act, where employees must remain in control, and where human judgment is part of the value proposition. They will treat agents neither as magic nor as toys, but as a new category of operational capacity that requires discipline.

That distinction matters for SMEs. Large companies can absorb failed experiments, reputational mistakes, and wasted implementation budgets. Smaller businesses often cannot. They have fewer layers of compliance, fewer technical specialists, and less room for error. But they also have an advantage. They can redesign work faster. They can decide what should remain personal. They can preserve culture deliberately.

This is why SMEs should not approach agents as a technology upgrade. They should approach them as an organisational design question.

Where does scale matter more than judgment? Where does consistency matter more than personality? Where does the founder need to retain control?

The trade-off is real. Agents can make a business faster, leaner, more responsive, more data-driven, and more productive. They can remove the administrative drag that suffocates smaller firms. They can professionalise reporting, sales preparation, customer support, and internal coordination. They can allow a small team to compete with larger incumbents.

But they can also change or even destroy the character of the company. Quietly. Gradually. One automated interaction at a time.

The future will not belong to businesses that automate everything. It will belong to businesses that decide, with precision, what should be automated and what should remain deliberately human.

The agentic economy is here. The choice is no longer whether software will work for us. It already does. The choice is who remains in control when it starts working on its own. If we delegate judgment entirely to software, the hierarchy flips: the agent becomes the operator, and you may become the tool.

By Louay Aldoory

Louay Aldoory is a Co-founder at 1648 Capital. 1648 Capital is a corporate advisory and private markets platform partnering with founders, shareholders, and investors on complex growth, restructuring, and capital structuring initiatives. We combine strategic insight with execution discipline, supporting businesses from transformation through to institutional capital alignment.

The strategies presented are thematic and do not constitute investment advice (or advice of any kind). No assurance can be given that the objectives of the investment above strategies will be achieved; the strategies involve risk (including, without limitation, illiquidity risk) and may incur a loss on some or all capital deployed. The opinions expressed, or indeed the information or assumptions that underpin them, may contain errors, mistakes, or omissions; no assurance or warranty can be made as to the accuracy or completeness of this information, and readers should not place any reliance on this content to execute investment decisions or for any other purpose. Readers accept full responsibility for using this content and are kindly requested to consult with their professional advisor before making any investment decision related to the same.

Change begins with a conversation