Autonomous Robot Risk in 2026: Is Your Business Ready?

Imagine that one of your competitors is operating their warehouse throughout the night. No supervisors on the floor at 2 am, no gap during shift change, and no backlog waiting for the morning shift. Its autonomous mobile robots are picking, sorting, and staging on an ongoing basis with a small team monitoring for exceptions on a dashboard. Down the road, another similar business still has to send all odd requests a manager's way, get their sign-off, and assume that no one is on vacation. Both companies are looking good on this quarter's numbers. Three years from now, only one of them will still be competitive.

​That is the real shape of the threat posed by the autonomous robot. It is easy to dismiss robotics as a manufacturing curiosity or a warehouse novelty. Still, the machines reshaping operations today are not the fixed-path, single-task robots of the last three decades. An autonomous robot perceives its environment, makes situational decisions, adapts to changing conditions, and coordinates with other systems with minimal human input. That distinction, between following instructions and interpreting a situation, is what separates old automation from the current wave.

​Most coverage of this shift asks a narrow question: will robots take your job? It is the wrong question for a business leader. The more consequential question is what happens to a company built around slow, sequential, human-dependent workflows when a competitor rebuilds its operations around continuous, self-correcting ones. This article works through that question in detail, across manufacturing, logistics, healthcare, retail, agriculture, construction, hospitality and office operations, and gives you a practical way to assess your own exposure, model the economics, understand the new risks, and plan an adoption path that does not overreach.

​The autonomous robot is most dangerous to businesses that mistake current profitability for future competitiveness.

Also Read: Agents vs Workflows: The Definitive 2026 Guide to Smarter AI Automation

​Is the Autonomous Robot a Threat to Your Business?

​The honest answer is that it depends. Threat level is not a property of an industry; it is a property of how a specific company operates within that industry. Six factors determine exposure.

• How repetitive the company's core workflows are.
• How much day-to-day operations depend on manual coordination between people.
• Whether direct competitors have already begun adopting robotics.
• Whether the business has clean, usable operational data to work from.
• How quickly management is able and willing to redesign existing processes.
• Whether the business competes primarily on speed, cost, accuracy or availability.

A company strong on all six can treat autonomous robots as a future opportunity to evaluate calmly. A company weak on most of them is closer to the edge than its balance sheet suggests.

Table 1 · Autonomous Robot Business Risk Snapshot

​Notice that the "lower risk" row is not a business with no robots; it is a business that has already absorbed the disruption and is now managing a different, more mature problem: integration complexity and governance, not survival. Every other row on this table is exposed to a competitor's decision, not just its own.

​What Makes an Autonomous Robot Truly Autonomous?

​The word "robot" covers a wide range of machines with very different risk profiles. A traditional industrial robot arm welding the same joint on a car body thousands of times a day is automated, but not autonomous: it has no awareness of its surroundings beyond a fixed set of sensors tuned to one task. Move the part six centimeters, and it fails. A collaborative robot, or "cobot," adds basic safety sensing so it can work near people, but its task logic is still largely fixed.

​Autonomous mobile robots, humanoid robots, delivery drones, robotic delivery vehicles and AI-powered inspection robots sit further along the spectrum. They combine perception, spatial reasoning and adaptive decision-making, so they can operate in environments that change, choose between multiple valid paths to a goal, and hand off to a person only when something falls outside their competence. Software automation, by contrast, has no physical body at all; it manipulates data and decisions rather than materials, but the reasoning layer underneath is often shared with physical autonomy.

​What actually produces autonomy is a stack of capabilities working together rather than any single sensor or algorithm.

​

Fig. 01 — The Autonomous Robot Intelligence Stack

​The practical takeaway for a business leader is this: a machine that follows a fixed path on a warehouse floor is not equivalent to one that reads the floor, reprioritizes its own task list when a route is blocked, and coordinates with three other robots to avoid a collision. Vendors will often use "autonomous" loosely. Before buying, ask specifically which layers of this stack the product actually implements, and which are marketing language.

​The Real Threat Is Workflow Replacement, Not Job Replacement

Most debates about robotics focus on whether a machine will replace an individual worker. That's the wrong frame of reference for the more disruptive structural change. The autonomous robots are not going to replace one task in an existing process; they may perform the entire process.

Typical workflow: Request received, request reviewed/approved by manager, employee assigned, manual work performed, some person reviews the work, corrects if necessary, report to manager.

​An autonomous workflow compresses that chain. A system detects that a task is needed, assigns a robot suited to it, the robot completes the task, the result is verified automatically, records updates in real time, and a human is only pulled in when something falls outside expected parameters.

Fig. 02 — Traditional Workflow vs Autonomous Workflow

​This compression changes more than throughput. It changes how many management layers are needed to keep work moving, how shifts are structured, where quality control happens, how inventory is planned, what gets reported and to whom, who is accountable when something goes wrong, and, perhaps most importantly, what customers come to expect as normal turnaround time. A business that has not examined its own sequence, only its individual tasks, will misjudge both the size of the opportunity and the size of the risk.

​Seven Warning Signs Your Workflows Are Vulnerable

​Use this as a short self-diagnosis rather than a scorecard to please a consultant. Each sign below reflects a workflow feature that autonomous robots exploit particularly well.

1. Tasks are repetitive and rules-based

Operational example

A picker follows the same picking pattern on each shift and uses the same packing rules, which are the same across all shifts.

Why robots outperform

The most reliable way to use autonomous systems is when the tasks are repetitive with clear rules.

Cost of ignoring it

Competitors automate the repeatable layer and divert workers to more valuable tasks, further widening the cost differential, which increases by the quarter.

First step

Record each step of the task on paper and in time; if it is possible to write it as a checklist, then it can probably be automated.

2. Work stops when specific employees are absent

Operational example

One experienced technician is the only person who can run a certain inspection, and output drops whenever they are out.

Why robots outperform

A robotic process does not call in sick, and knowledge encoded in software does not walk out the door with an employee.

Cost of ignoring it

Single points of human failure quietly cap how much the business can grow.

First step

Document the tacit knowledge behind the bottleneck before considering any automation of it.

3. Managers spend significant time assigning routine tasks

Operational example

A standard job has to be manually distributed by a shift supervisor in the first hour of each working day.

Why robots outperform

Task orchestration software can match work to capacity during the shift, as well as at the beginning of the shift or once the shift has begun.

Cost of ignoring it

Instead of the decisions only managers can make, skilled managers are used up on logistics.

First step

Measure the amount of work a manager does that is not "assignment" work and the amount of work that is "judgment" work.

​4. During peak traffic, mistakes are likely to occur.

Operational example

During the busy season, the accuracy of the order is significantly reduced due to tiredness and stress among employees.

Why robots outperform

A well-tuned autonomous system performs at a consistent standard regardless of volume, within its designed capacity.

Cost of ignoring it

Peak-season errors are disproportionately expensive because they land on your highest-value customers.

First step

Track error rate against volume, not just in aggregate, to see the real curve.

5. Operations depend on spreadsheets, calls, or paper forms

Operational example

Stock levels are reconciled from a paper count sheet at the end of each day.

Why robots outperform

Autonomous systems need structured, real-time data; paper-based operations cannot connect to them at all without a digital layer first.

Cost of ignoring it

The business is not just behind on robotics; it is behind on the basic digital foundation robotics requires.

First step

Digitize the underlying record before evaluating any physical automation.

6. Customers frequently wait for internal handoffs

Operational example

An order sits between departments waiting for one team to notify the next.

Why robots outperform

Replacing workers does not eliminate the delay; automating the handoff and redesigning the layer does.

Cost of ignoring it

Delay is a company issue, not a department issue, and a customer problem; they go elsewhere.

First step

Map every handoff in the order-to-delivery chain and measure the time lost at each one.

​7. Competitors are offering faster or cheaper service through automation

Operational example

A rival now advertises same-day fulfillment that your current process cannot match.

Why robots outperform

Once one competitor resets customer expectations, the whole category is judged against that new baseline.

Cost of ignoring it

Margins erode as the business is forced to discount to compensate for slower service.

First step

Benchmark your cycle time directly against the fastest competitor in your category, not the average one.

​

​Which Traditional Workflows Will Break First?

​It is more useful to think in terms of workflow categories than individual job titles, because a single job often contains several workflows with very different exposure levels. Material movement, picking and packing, inventory counting, visual inspection, cleaning and sanitation, security patrols, last-mile delivery, agricultural monitoring, repetitive assembly, internal document or parcel movement, routine patient support, and hazardous-environment inspection are the categories currently seeing the fastest adoption of autonomous systems.

Table 2 · Workflows Most Exposed to Autonomous Robots

​High exposure doesn't necessarily mean immediate replacement everywhere. Other factors that influence adoption rates are environmental complexity, maturity of relevant regulation, safety requirements, and whether or not the economics work for a specific site. If the physical space is exceptionally busy or chaotic, then in principle a workflow can be at "very high" risk of disruption, but take years to actually have this disruption happen. The table represents a map of pressure, NOT a countdown clock.

​What the table does make clear is that the categories under the most pressure share a common feature: they are structured, repeatable, and measurable. Any workflow you can already describe as a checklist is one a vendor can already describe as a product.

​Industry-by-Industry Impact of Autonomous Robots

​Manufacturing

Manufacturing already has the deepest robotics history, and the newest layer is autonomy rather than automation itself. Autonomous assembly cells that reconfigure for different products, machine tending robots that load and unload equipment without a fixed script, camera-based defect detection, predictive maintenance that flags failures before they happen, internal material transport between stations, and, in a small but growing number of facilities, lights-out manufacturing that runs with no people on the floor at all are all in active deployment.

​Warehousing and Logistics

This is the sector where the business case is currently clearest. Autonomous mobile robots, dynamic picking that adjusts routes in real time, automated sorting, continuous inventory scanning, robotic loading and unloading, and last-mile delivery robots and drones combine into a fulfillment operation that can run around the clock and rebalance itself as order volume shifts through the day.

​Healthcare

In healthcare, the most rapidly growing robotics applications are in support functions such as medicine delivery within hospitals, internal logistics between departments, continuous patient monitoring, disinfection robots, rehabilitation assistance devices, and more and more sophisticated surgical assistance tools, which augment the surgeon's precision, rather than replace it. To be specific, the immediate impact is in logistics and support – not in autonomous clinical decision-making, which is tightly controlled and appropriately overseen by people.

​Retail and Hospitality

Shelf-scanning robots that flag out-of-stock items, automated stock replenishment, floor cleaning robots, food delivery robots on campuses and in dense urban areas, customer guidance kiosks, and semi-automated kitchen systems are changing both the cost structure and the customer experience in these sectors simultaneously.

​Agriculture

Even if technical maturity is limited by crop type, the economic case is compelling to precision spray individual plants rather than groups; allow the harvesting of individual plants for some crops; use camera technology to detect weeds; and monitor livestock using sensors; all of these are solutions that address a problem faced by many farmers in the sector for years—seasonal labor deficits.

​Construction and Mining

Autonomous vehicles for earthmoving and haulage, drone- and robot-based site mapping, structural inspection in locations too dangerous for people, hazardous material handling, and remote operation of heavy equipment are advancing steadily, constrained mainly by the unpredictability of physical sites rather than by the technology itself.

​Offices and Professional Services

The introduction of physical robotics will be more gradual in offices than in warehouses or factories for a number of reasons, including the lack of structure in an office environment, as well as the economic case. However, the much-anticipated embodied AI – robots and software agents that have both a physical presence and a cognitive mind – will begin to assume reception duties, building security, document and parcel delivery, and facilities management tasks, in addition to the current transformation of administrative tasks with digital-only automation. That's because office leaders who think robots are “not their problem” are likely the ones most vulnerable to a more muted and, at the moment, less flashy kind of disruption hitting them first via software.

When Does an Autonomous Robot Become Cheaper Than Labor?

There is no comparison between an employee's salary and a robot's purchase price. A true cost model should be built to incorporate the costs of the hardware, software subscriptions, integration efforts, site preparation, maintenance costs, connectivity, cybersecurity measures, insurance, staff training, downtime, energy usage, redevelopment of the surrounding process, and hardware depreciation. The human alternative is the whole cost, which includes salary, overtime, recruitment, training, staff turnover, workplace injury, human error, loss of productivity because of staff unavailability, management overhead, etc.

Table 3 · Human Workflow vs Autonomous Robot Cost Model

​Robot break-even period = Total implementation cost ÷ Annual net operating savings

​Do not calculate return on investment by comparing a single robot's price against a single employee's salary; that comparison almost always understates the real picture in both directions. A robot's savings usually come from several roles at once, fewer errors across the process, higher throughput, and reduced downtime, all of which need to be measured at the workflow level, not the headcount level.

​Productivity Gains Can Hide New Operational Risks

​Efficiency numbers and failure modes are the way the automation press reports. A complete list would also include robots that perform poorly in unstructured environments, sensor inaccuracies in poor lighting or unusual conditions, poor handling of edge cases that were never a part of the system's training, reliance on connectivity that can fail at the most inopportune moments, integrations that fall apart because of incompatibility between old and new systems, vendor lock-in that precludes future flexibility, maintenance delays when a master technician is unavailable, employee resistance that hinders the adoption of new processes, safety issues, poor AI perception, poor escalation system, or simply over-automating a process that required a human decision to be put in the middle of it.

​These failure modes point to a broader concept worth naming directly: automation fragility. A highly automated company can be extremely efficient under normal operating conditions and unusually vulnerable the moment a central platform, network connection, or single vendor fails. Efficiency and resilience are not the same property, and a business can optimize hard for one while quietly eroding the other.

​

Fig. 03 — The Autonomous Robot Risk Web

​Cybersecurity: When a Robot Becomes a Physical Attack Surface

​An autonomous robot is not just a piece of equipment; it is a networked computer that happens to move through physical space. It combines cameras and sensors, business data, wireless connectivity, cloud platforms, AI decision systems, and integrations with enterprise software, which means it inherits every conventional cybersecurity risk and adds a physical dimension on top.

​The realistic threat list includes remote hijacking of a robot's controls, spoofed sensor input that fools its perception, theft of mapping data that reveals a facility's layout, compromised credentials, malicious software updates pushed through a compromised vendor channel, attacks that spread across an entire fleet at once, vulnerabilities introduced through the supply chain, surveillance and privacy concerns from always-on cameras, and ransomware that can halt physical operations rather than just locking files.

​The difference between a compromised laptop and compromised robot is that — with a compromised laptop, a data breach will likely result; with a compromised robot, a physical safety incident will likely result. This changes the level of seriousness and makes robot cybersecurity a broad-level concern, not just an IT issue.

Reasonable controls include network segmentation – a robot can't reach systems that are not connected; zero-trust security policies – software and software updates are cryptographically signed; detailed activity logging – logs are kept of who accesses the system, what they do with it, and how they access it; clear emergency stop procedures – there are clear procedures for when things go wrong; a reliable human override – clear and well maintained procedures for when a human needs to override the system; formal vendor security assessments – there are formal assessments of the security of the vendor's systems; and regular penetration testing – identify weaknesses before operational failures expose them.

Which is More Important: Safety or Liability?

At some point in every deployment, there are questions no one wants to answer when they are being deployed. Whose fault is it if the robot hurts somebody? Who is responsible for the liability: the manufacturer of the machine, the software provider, the systems integrator, or the business that will be using the machine? Decisions are made by a system that continues to learn after deployment, so who approves the decisions? What is the result if a robot executes its program word for word and still results in an unsafe outcome? How do you need to store evidence from incidents, and when is it necessary for a person to assume direct control?

These questions relate to workplace safety legislation, product liability, insurance policy, formal risk assessment, human override design, audit trails, restricted operating zones, reporting of incidents and the governance of the continued software and model updates. It's none of these exotic things, but all of these need an owner before – not after – an incident.

Table 4 · Autonomous Robot Governance Responsibilities

​What is noteworthy is that the pattern "Business Operator" is the primary owner, as its name implies, more often than any other party, such as workflow design and access control. Do not outsource responsibility for the technology. Before you sign any deployment contract, name each of these rows in this table, but after the first incident, it is too late.

​How Autonomous Robots Reshape Management and Organizational Structure

​Robotics removes coordination work as well as physical work, and that second effect is easy to underestimate. Fewer people are needed purely to assign routine tasks. At the same time, demand rises for fleet-management roles, robotics technicians, data engineers who keep the underlying information clean, and new automation-governance functions that did not previously exist. Managers increasingly oversee outcomes rather than individual tasks, monitor performance continuously rather than at scheduled intervals, and work across a blended IT-and-operations remit that used to sit in two separate departments.

​This adds up to what is best described as a robotic operating model: people manage goals, constraints and exceptions, while autonomous systems handle the bulk of execution.

​

Fig. 04 — Traditional Organization vs Robotic Operating Model

​Jobs Will Change, but Which Human Capabilities Become More Valuable?

​Reducing this to "robots will take jobs" flattens a more useful picture. Roles are best understood in three groups.

Tasks Likely to Decline

• ​Repetitive movement
• Routine inspection
• Manual stock counting
• Predictable transport
• Basic machine tending
• Repetitive cleaning

​Roles Likely to Grow

• Robotics maintenance
• Fleet supervision
• Automation engineering
• Safety governance
• Cybersecurity
• Data analysis
• Process design
• Human-machine interaction

​Hard to Replace

• ​Complex judgement
• Empathy
• Negotiation
• Ethical accountability
• Creativity
• Leadership
• Unstructured problem-solving

Handling unfamiliar environments

​The strategic implication is Sequencing. Change roles before cutting people out of the organization, not after. Operational knowledge is a capability that only comes with years of experience, and no robot vendor or systems integrator can offer on day one; and it's a knowledge that is best to retain as late as possible in an automation program, if at all.

​Build, Buy or Subscribe: Choosing the Right Robot Adoption Model

​Direct purchase suits mature, predictable, high-volume processes where the workflow is unlikely to change significantly for several years. Robotics as a service suits organizations that want a lower initial outlay, flexible capacity, and vendor-managed maintenance, which is often the right entry point for a first deployment. Custom development is reserved for strategically important workflows that no standard product on the market can serve.

Table 5 · Autonomous Robot Adoption Models

​The Autonomous Robot Readiness Audit

​Score your organization from 1 to 5 across six dimensions, for a maximum of 30 points.

• Workflow suitability: are processes repetitive, measurable, and stable?
• Infrastructure readiness: does the site have suitable layouts, connectivity, power and safety controls?
• Data readiness: are there accurate and readily available operational data?
• Integration readiness: can robots connect with warehouse, ERP, production or customer systems?
• Workforce readiness: are employees informed, trained and involved?
• Governance readiness: are accountability, safety, cybersecurity and escalation procedures defined?

​

​

Fig. 05 — Autonomous Robot Readiness Radar

​A 90-Day Autonomous Robot Pilot Plan

​Days 1 to 15: Identify the Right Workflow

Do an end-to-end walkthrough of the current process and honestly measure cycle time; calculate the existing error rate; document the labor and downtime cost; find out if there are safety constraints; and choose one narrow use case instead of multiple use cases.

Days 16 to 30: Define the Business Case

Establish clear performance goals, determine total cost of full rollout, define what success and failure look like in writing prior to the pilot, include the frontline staff who will be impacted, and narrow down the list to 2-3 vendors.

​Days 31 to 60: Deploy a Controlled Pilot

Deliberately restrict the operating domain, always have a manual option, train supervisors in depth and not just for a few hours, test edge cases intentionally, have a very close eye on safety, and gather operational data throughout the pilot, rather than waiting for final evaluation.

​Days 61 to 75: Evaluate Results

Measure throughput, error reduction, downtime, how often employees had to intervene, safety incidents, cost per task, and the actual impact on customers.

​Days 76 to 90: Decide Whether to Scale

Choose one of four honest outcomes: scale immediately, modify and retest, move to a different workflow, or stop the project. All four are legitimate outcomes of a well-run pilot.

​

Fig. 06 — Autonomous Robot Adoption Decision Path

​What Happens When Competitors Adopt Autonomous Robots First?

​Early adopters in a category are likely to benefit from quicker fulfillment, greater operational uptime and more effective operations, lower variable costs per item, more standardized production, real-time inventory visibility, reduced lead times, more personalized production, and increased resilience to labor shortages. All of these benefits are real and are being seen in logistics, food delivery, and precision agriculture today.

​The mechanism that anybody can call is the expectation reset effect. When, after a first experience of faster, cheaper, or continuously available service from one provider, the usual delivery times of all the other providers of the same type appear to become intolerable, even when they haven't made any changes whatsoever. The widespread adoption of a few early adopters is driving a category-wide standard shift, such as same-day fulfillment as an expectation, shrinking production lead times, high standards for consistency with robotic inspection, new food-delivery business models with automated kitchens, and control over input usage with agricultural robots.

​When Not to Deploy an Autonomous Robot

​Robotics is not the right answer everywhere, and pretending otherwise damages credibility as much as ignoring the technology does. Robots tend to be a poor fit when the workflow changes every day, the environment is highly unpredictable, human interaction is central to the value the customer is actually paying for, the underlying process has never been standardised, task volume is too low to justify the investment, safety risk cannot be adequately controlled, the company lacks the technical support to maintain the system, data privacy requirements cannot be satisfied, a simpler and cheaper automation technology solves the same problem, or, most damaging of all, when management is pursuing robotics mainly as a publicity exercise rather than an operational one.

​Automating a poorly designed workflow does not fix the workflow. It makes the same inefficiency happen faster and at greater expense.

​The Hybrid Workforce: The Most Realistic Future

​The most credible outcome across almost every sector is collaboration rather than complete replacement. Autonomous robots are well suited to repetition, physical precision, continuous monitoring, heavy transport, hazardous tasks, and structured decision-making. People remain essential for handling exceptions, managing relationships, making strategic decisions, carrying ethical responsibility, improving processes, solving unfamiliar problems, and operating in conditions nobody anticipated.

​This is the practical meaning of human-on-the-loop operations: workers do not directly control every individual action a system takes, but they monitor its performance continuously and intervene the moment something falls outside expected bounds.

Fig. 07 — Human–Robot Collaboration Loop

​The Final Verdict: Threat, Opportunity or Both?

​Return to the question in the title with a clear answer. Autonomous robots genuinely threaten workflows that are inefficient and inflexible. They do not automatically threaten every employee or every business, and treating this as an all-or-nothing prediction misreads how the technology actually spreads through an industry. Companies that redesign work around human-machine collaboration stand to gain real advantage. Companies that delay experimentation altogether are more likely to face rising costs, slower service, and a shrinking competitive position, quietly, over several years rather than all at once.

​The first step is not purchasing a robot. It is identifying, honestly and specifically, where your organization is currently losing time, quality, and money. Everything in this article- the risk snapshot, the readiness audit, the pilot plan- only becomes useful once that first step has been done properly.

​The autonomous robot will not destroy every traditional business. It will expose which traditional businesses were already too slow to survive the next operating model.

​Frequently Asked Questions

​What is an autonomous robot?

​An autonomous robot is a machine that perceives its environment through sensors, reasons about what it observes, and carries out physical tasks with minimal ongoing human instruction. Unlike a fixed-path machine, it can adapt its actions as conditions change and escalate to a human only when needed.

​How is an autonomous robot different from a regular robot?

​A regular industrial robot repeats a fixed programmed sequence and fails if conditions shift even slightly. An autonomous robot senses its surroundings, adjusts its plan in real time, and can handle variation, obstacles, and new situations that were never explicitly programmed.

​Will autonomous robots replace human workers?

​Particular repetitive tasks are more likely to be replaced by robots than jobs. There will be change to most roles rather than elimination – much of the repetitive work will be automated, and human involvement in the role will move to judgment, exceptions, and relationship work.

​Which industries will autonomous robots affect first?

​Manufacturing, warehousing, and logistics are furthest along, followed closely by agriculture and parts of healthcare logistics. Retail and construction are adopting more selectively, while office and professional-service environments will likely see the slowest and most gradual physical impact.

​Are autonomous robots expensive?

​Hardware is only one line item. A realistic budget includes integration, site preparation, maintenance, connectivity, cybersecurity, and staff training. Robotics-as-a-service subscription models can substantially lower the upfront cost for businesses not ready for a full capital purchase.

​Can small businesses use autonomous robots?

​Yes. Another way for smaller businesses to gain access to autonomous capabilities, without the need to purchase the robots themselves, is through robotics-as-a-service and outsourced, automated fulfillment providers that operate with a pay-as-you-go or managed-as-a-service pricing model.

​What are the main risks of autonomous robots?

​The core risks are physical safety, cybersecurity exposure, unplanned downtime, unclear legal liability, workforce disruption, and heavy dependence on a single vendor. These are not the reasons to avoid the technology outright, but each needs a named owner before deployment.

​Are Internet-enabled autonomous robots necessary?

Depending on the system, onboard processing may be all that is required for some core navigation and safety functions. In contrast, more sophisticated reasoning, fleet coordination and remote monitoring may require cloud connectivity. Businesses should ask vendors about which functions are lost when the system is disconnected.

What does a company need to do in preparation for autonomous robots?

Map existing processes in detail, clean up as much operational data as possible before implementing, and engage employees early, not late; establish governance and accountability before deployment, and conduct a small, well-specified pilot project before committing to a wide rollout.

Does it always make more sense to have an autonomous robot rather than a human?

No. Robots work best in repetitive (measured), structured (predictable) environments. Most realistic deployments are a mixture of human workers and autonomous systems; humans are stronger in situations where they are not used to, where judgment is uncertain, or in situations involving relationships.

​