- Competitors with fewer people don't win because they had the nerve to cut staff. They win because they designed the work so each person produces more.
- The numbers to watch are revenue per employee and the cost of each unit of work. Headcount on its own tells you little.
- Start by measuring your own real processes. Cutting staff doesn't need to be on the table yet.
1. The competitive picture for 2028 to 2029
“Half the headcount” sounds scarier than it is. It doesn't mean every company has to lay off half its staff. It means that a piece of work your company needs people to touch six times, a competitor may handle with a single touch. They can put that difference into lower prices or into answering customers faster than you do.
Over the next two to three years, AI tools are likely to shift from assistants that wait for instructions to systems that take on continuous work, call on several kinds of data and tools, and pass only the exceptions to people. Companies that have designed their workflows in advance will be able to grow volume while adding people more slowly than revenue. Traditional companies will still need to hire more admin staff, coordinators and managers every time sales grow.
“Half the headcount” doesn't mean every company has to cut by half. It reflects a difference in how much human touch each transaction needs. Work that used to require someone to open the email, copy data, create documents, request approval and update systems may end up needing a person to review only 10 to 30 percent of items. If a competitor gets there first, it can use the difference to lower prices, add services or invest in winning new customers.
2. The competitive gap will come from systems, not tool subscriptions
It's normal for two companies that buy the same tool to get different results. The first has employees ask AI one question at a time and copy the answers back into the system by hand. The other sets things up so data flows between systems by itself. The difference lies in how the work is organized.
Two companies can use the same AI model and get different results. In the first, employees ask AI questions one at a time and copy the answers back into the system. The second has central data, business rules, workflows and monitoring, so AI can take on repeat work all day long. What separates them is the organization's operating system. Better prompts have little to do with it.
The gap compounds like interest. Once a workflow is running, the company collects more feedback and exceptions, the system improves, costs fall and the team has time to design the next process. Companies still experimenting in scattered ways have no shared data to learn from. Every department starts from zero and nobody dares hand real work to AI.
The other side is decision speed. If reports are summarized automatically and problems are flagged immediately, executives can adjust prices, stock or campaigns sooner. So the advantage includes fewer losses from spotting problems late, on top of lower labor costs.
3. Traditional companies compared with AI-enabled companies
| Area | Traditional company | Company designed around AI |
|---|---|---|
| When volume grows | Add people and team leads | Add system capacity and people to handle exceptions |
| Customer response | Depends on the queue and business hours | Standard questions answered instantly, hard cases sent to people |
| Documents | Created and checked by hand every time | Generated from central data and spot-checked |
| New products | Wait on several departments and a large budget | Prototype quickly and test with a small group |
| Managers | Hand out work and chase status | Oversee quality and exceptions, and adjust the rules |
An AI-enabled company doesn't need the smallest possible headcount. It chooses to put people on work that builds trust, requires judgment and produces new ideas. As long as the customer experience isn't diminished, a lower cost base is a more lasting advantage than cutting staff by the same amount in every department.
4. Assess whether your company is at risk of falling behind
Choose three core processes, such as Lead-to-Quote, Order-to-Cash and Issue-to-Resolution, and ask: how many times is the same data entered, how many handoffs are there, what percentage of the work is standard cases, and how many minutes of real judgment does an employee actually need? If more than half the time goes to searching, copying, summarizing and passing things on, there is plenty of room to automate.
Check your data systems too. If product lists, prices, contracts and customer history live in personal files, AI will struggle to connect the work. Companies that organize their data and access rights now will be able to adopt new models faster as the technology develops.
- A 10 percent increase in revenue requires almost 10 percent more staff
- Managers spend more than one day a week compiling reports and chasing work
- Customers wait for information that is already in the system
- There are several AI projects but no figures showing that unit costs have actually fallen
- There are no data owners and no register of AI systems
5. Five things a business should start adjusting today
- Set a process baseline: measure volume, touch time, wait time, errors and cost per task for your key workflows
- Create a source of truth: define product, customer, policy and pricing data, each with an owner and an update cycle
- Automate one path completely: instead of spreading chatbots across every department, start with a workflow that has an ROI and results you can verify
- Design human exceptions: decide which cases AI handles, which cases people take, and what the SLA is
- Build lasting measurement: the dashboard has to show quality, cost, speed and customer outcomes, rather than just the number of prompts
This preparation keeps its value as the technology changes, because process, data and governance carry over to many generations of models. A company shouldn't wait for AI to be perfect before organizing its data. By then, competitors with the foundations in place will have expanded several times over.
6. Plan your workforce without panic and without waiting too long
Classify roles by how much of their work falls into Automate, Augment and Human-led. There is no need for immediate layoffs. Leave vacancies unfilled, move people into revenue-generating work and train team leads to manage the system's queues. Only when the data shows a lot of spare capacity should you make structural decisions, and then make them transparently.
The skills worth investing in are process design, quality checking, working with data, deciding exceptions, caring for customers and turning front-line knowledge into rules a system can use. Employees who understand the business and can direct AI will have far more leverage than those who simply do manual work quickly.
Define three workforce scenarios, in which AI raises efficiency by 20, 40 and 60 percent, then calculate headcount, new roles and reskilling time for each. This keeps the company from making long-term hires into roles that may shrink before the real results are visible.
7. Invest without chasing every trend
For the development team · Technical detail
Use a portfolio split into Core Efficiency, Growth and Future Options. Core Efficiency has a clear payback target. Growth is tied to leads, conversion or new products. Future Options gets a limited experimental budget and a decision date. Every project needs a Process Owner and Exit Criteria.
For the development team · Technical detail
Avoid large platform projects before you have a proven workflow, but don't build separate tools that can't connect to each other either. Set shared standards for Identity, Data Access, Evaluation, Logging and Cost Control, then let teams build use cases on the same foundation.
8. A three-year roadmap you can adjust
Year one: measure the core workflows, put 2 to 3 use cases into production, and organize your data and security standards. The goal is proven business results and a core team that can repeat them.
Year two: connect workflows across departments, add specialized agents, and adjust roles and capacity planning to fit automated work. The goal is a clear rise in revenue per employee and in the speed of launching products.
Year three: rework the operating model, pricing and the offers that AI makes possible, such as personalized services, real-time responses or new low-cost products. Review the company structure based on actual results, without treating the old workforce plan as a constraint.
In short: A competitor running on half the headcount may owe its lead less to deeper cuts than to systems that let every person produce more. Businesses should start with process, data and measurement, while setting out workforce scenarios and investing as a portfolio. Starting today buys time to learn before pressure on price and speed becomes urgent.
OpenAI: How frontier firms are pulling ahead
Turn a cost advantage into a system you can measure
The knowledge of where your business loses time sits with your heads of sales and accounting and with the front-line staff who do that work every day. A software company doesn't have it. Before writing any code, DNA Maker sits down with those teams and traces the real workflow, from the moment a customer gets in touch to the day the money is collected: how many times data is entered again, how many handoffs there are, which items are standard cases and which always need a person to decide. The output of this stage is a process map, rather than a sales quote, that lets the owner see their own unit costs more clearly.
What we help you build next
From that map, we design a system that connects central data to a single workflow, so the system handles standard work continuously and sends only the exceptions to a queue for people, with a dashboard that shows executives the time from receiving work to delivering it and the cost per transaction every week. We always start with one workflow that can be measured, build a prototype for the team to try in real use, measure it against the baseline, and only then expand. If you have a process you already know eats time but haven't dared to touch, let us take one look at its real data.
SOFTWARE ENGINEERING GLOSSARY
Software engineering glossary
The terms in this article are about measuring processes and connecting systems. Read the questions on the right and put them to your development team before you approve a budget.
| Term | What it is | A simple example | What executives should ask the development team |
|---|---|---|---|
| Workflow | The sequence of work from start to result, naming who is responsible and the conditions at each step | From a customer asking for a price to issuing the quotation and approving the discount | Who does this workflow start and end with, and where is the time measured? |
| Handoff | A point where work passes from one person or team to another, often where queues form and information gets lost | Sales sends details to accounting to work out a price, then waits for it to come back | How many handoffs can we remove, and how long do the remaining ones take? |
| Cycle Time | The total time from when work arrives to when it is delivered, counting the waiting as well as the hands-on time | A quotation takes 40 minutes of actual work, but the cycle time is 2 days | Is cycle time measured automatically by the system, or do people enter it by hand? |
| Baseline | The starting figures before a project begins, used to check whether things have really improved | The cost per quotation in the month before the system went live | Without a baseline, how will we know whether it was worth it? |
| System Integration | Connecting existing systems so they pass data to each other without anyone retyping it | Connecting the CRM to the accounting system so prices match automatically | How can our existing systems be connected, and what are the options if they can't be? |