Business resilience & company knowledge

What would your business be worth if your best people left tomorrow?

What would your business be worth if your best people left tomorrow?

Build a business that retains its knowledge, supports its people and can keep performing when key individuals are unavailable.

CSM · Video · 12 min read · Updated 18 November 2026

The short answer

Start by identifying the work that only one person can complete. Capture the knowledge and evidence behind it, agree authoritative data sources, and build clear handovers, controls and cover. AI can help prepare and coordinate the work within those systems. Measure whether the business becomes more dependable; a valuation increase is not automatic.

Start by identifying the work that only one person can complete. Capture the knowledge and evidence behind it, agree authoritative data sources, and build clear handovers, controls and cover. AI can help prepare and coordinate the work within those systems. Measure whether the business becomes more dependable; a valuation increase is not automatic.

Watch: What would your business be worth if your best people left tomorrow?

Watch how CSM approaches company knowledge, connected systems and business resilience.

What would your business be worth if your best people left tomorrow?

You cannot calculate the answer from a single scenario. You can use the question to expose where the business depends on knowledge, relationships and decisions that have never become shared capability. If quoting, delivery or financial reporting stops when one person is unavailable, that dependency deserves attention before you add more work or more technology.

A useful first test is practical: could another authorised colleague complete an important job using the information and systems available, without calling the usual expert? Record what they cannot find, understand or approve. Those gaps give you a starting point for reducing risk and making good work repeatable.

What is key-person dependency?

Key-person dependency exists when an essential outcome relies on one individual’s availability, memory, access or relationships. It can sit with a founder, a long-serving estimator, an operations manager or the only person who understands a spreadsheet. Seniority is less important than what stops when that person is absent.

For example, a sales team might have a CRM and still depend on one director to explain which promises were made to a customer. Finance might have current invoices but need one colleague to interpret unusual adjustments. Having software does not mean the reasoning required to use it is shared.

The aim is to make expertise available to more people while keeping responsibility clear. Experienced people remain valuable for their judgement, relationships and ability to handle unfamiliar situations. A stronger system reduces the routine demands on their memory and attention.

Why does this matter to business value?

An investor or buyer needs confidence that earnings and service can continue after a change of ownership or personnel. Dependable records, clear responsibilities and a capable management team help explain how the business operates and what would transfer. They also help a leader grow the company without becoming the approval point for every decision.

BDC identifies transferability, including the ability to operate without the current owner, as a factor in a business sale. Strong management, documented processes and clear systems support that transferability. Read BDC’s guidance on preparing for sale.

There is no automatic valuation uplift from installing AI. Earnings quality, customer concentration, contracts, market conditions and many other factors affect value. CSM’s role is to improve the operating capability underneath the business: the information, processes and decisions that help it perform consistently. The financial effect needs to be evidenced in that particular business.

Where should you look for hidden dependency?

Follow one real piece of work from its first request to its finished result. Look for the points where people search, retype, wait and check. Ask who resolves the uncertainty and whether the next person could do the same job using an approved record.

Pay attention to workarounds: a private inbox holding customer commitments, a locally saved rate card, a shared login, or an undocumented rule for accepting an order. Ask what happens during annual leave, when someone changes role or when two urgent jobs arrive together. These scenarios often reveal the difference between a documented process and a workable one.

Record the business impact as well as the delay. An infrequent approval may protect a major commercial commitment. A five-minute task can carry more risk than a daily report. Do not let a combined score hide a dependency that could interrupt delivery or expose the company to an unacceptable commitment.

How do you turn individual knowledge into company capability?

Capture the reasoning behind a real decision, not just the final answer. What information did the expert use? Which version was current? Which exceptions mattered? Why did they choose one action over another? Turn that explanation into a playbook with examples, acceptance criteria and a route for escalation.

Give the playbook an owner. Distinguish a general company rule from a correction that applies only to one customer or project. Review changes before sharing them, retain source references and restrict access to confidential information. Every chat message should not silently become a new rule.

Then test whether another person can use the playbook to complete representative work. If they still have to ask the expert the same questions, improve the instructions, evidence or handover. A useful knowledge system needs regular review, practical training and cover for the people who maintain it.

Why does becoming data-first come before dependable AI?

AI needs reliable information to prepare useful work. If nobody agrees which scope, price, customer record or drawing revision is authoritative, adding a model will not resolve that disagreement. Data-first means assigning ownership, agreeing definitions and making the right information available with the right access.

Start with the data required for one priority workflow. Decide which system owns each record, how records are matched and how changes are checked. A trusted source of truth does not require copying everything into a single database. It requires an agreed authoritative source for each important fact and a dependable way to use it.

Connections also need monitoring. Who notices when a feed stops updating? How are duplicates and conflicting versions handled? Can access be revoked? CSM’s existing integration experience can provide a starting point, but each implementation still needs the appropriate access, mapping, permissions and tests.

Where should people, systems and AI each contribute?

People set direction, own relationships and make the decisions that require judgement and authority. Systems provide records, responsibilities, access rules and repeatable handovers. AI can help gather relevant information, interpret variable inputs, prepare drafts and follow up within agreed boundaries.

For a quotation, the system might assemble customer requirements and approved rates. AI could prepare a scope and identify missing information. A qualified person would resolve ambiguity and approve the commercial commitment. Controlled rules can apply approved prices without asking a model to invent them.

Use ordinary automation where the rule is explicit. Add an agent where the task needs interpretation or a choice of actions, and only give it the tools and permissions required. Anthropic’s guidance recommends starting with the simplest approach that meets the need and distinguishes predefined workflows from agents that direct their actions dynamically. Read the agent-design guidance.

What are the six layers in CSM’s approach?

The six layers describe how we connect useful information to useful action. They are an implementation framework, not a requirement to replace every system or purchase six separate products. Leadership sets priorities first; the build then starts with the information needed to deliver them.

1. Data: establish the authoritative records, definitions, owners and connections needed for the chosen work. Test completeness, freshness and permissions before relying on the information.

2. An AI operating layer: give people a shared place to see work, retrieve approved information, review outputs and understand what needs attention. Define responsibilities and access around the workflow.

3. Knowledge, skills and playbooks: record how the business performs the task, the evidence it uses, the checks required and when a person must intervene. Keep that knowledge maintained and versioned.

4. Models: choose and test a suitable model for each task against representative examples. Keep company knowledge and evaluation criteria usable as models and suppliers change.

5. Agents: assign bounded jobs, expected outputs, permitted tools and escalation routes. Improve performance through reviewed feedback and testing. More autonomy should follow evidence, rather than an assumption that an agent will improve automatically.

6. Commercial intelligence: connect operational evidence to the decisions leaders need to make. Surface constraints, exceptions and exposure with enough context to decide what action is appropriate.

What does commercial intelligence help a leader decide?

A useful view explains what needs attention, why it matters and who can act. A traffic-light status is only meaningful when its definition, source, freshness and threshold are understood. Leaders should be able to inspect the underlying evidence rather than accept an unexplained red, amber or green label.

For example, a pipeline total can look healthy while approved quotes are waiting for one specialist. Bringing those facts together may show that the immediate constraint is review capacity. The useful response could be better evidence at handover, clearer approval limits or an additional trained reviewer. Faster drafting alone might increase the queue.

Agree what should happen when an alert appears. Name the responsible person, the evidence they receive and the action or decision expected. Commercial intelligence earns its place by helping people intervene earlier and follow through, not by creating another report to read.

What are the seven steps from discovery to live use?

1. Discover the business and the work. Agree the leadership priorities, examine real examples with the people involved and map where information or decisions get stuck. Capture dependencies and operational risks as well as potential time savings.

2. Prioritise and scope. Use the scorecard to assess commercial impact, capacity, key-person dependence, the risk of inaction and readiness. Turn the selected priorities into a ninety-day roadmap with owners, boundaries and measures of success. Treat serious risks explicitly, even if their numerical score is low.

3. Prototype. Use appropriately authorised real examples to test the proposed workflow before the full build. Show the team the inputs, outputs, approvals and exceptions. A prototype tests assumptions; it does not establish production readiness.

4. Build and test. Implement the agreed scope and test difficult cases: missing records, conflicting instructions, changed permissions and failed connections. Verify that consequential actions cannot bypass their approval requirements.

5. Pilot with the team. Let users complete representative work, challenge outputs and report problems. Compare the whole workflow with the baseline, including human review and correction time. Resolve gaps against the agreed acceptance criteria.

6. Go live with support. Agree ownership, monitoring, recovery and support responsibilities. Train people on the jobs they actually perform, including how to correct an error, raise a concern and stop or escalate the workflow.

7. Review and repeat. Assess the live result, update the scorecard and select the next priority. Revisit the roadmap every ninety days and build on what works. The roadmap is a planning cadence, not a promise that every system will be delivered within ninety days.

How do you know whether the investment is working?

Measure outcomes before and after the change using comparable work. Useful measures include turnaround time, rework, approval waiting time, missed handovers, repeat errors and the number of tasks that only one person can complete. Include software costs, support and the effort needed to review AI outputs.

Time released becomes financially valuable through what the business does with it. It might create capacity for additional profitable work, avoid an otherwise necessary hire or reduce a genuine expense. Those are different outcomes. Hours saved do not automatically become cash savings or an improvement in EBITDA.

Make the business case explicit. Estimate implementation and ongoing costs, the likely adoption level and the benefit that can reasonably be realised. Review the assumptions after the pilot. A payback period should follow that calculation, rather than a universal promise that an AI project pays for itself within a few months.

What should you agree about ownership, guarantees and support?

Ask what you will receive and what you can continue using if the relationship changes. The agreement should specify ownership and access for custom code, data, documentation, workflows and configuration, alongside any third-party licences, hosting and ongoing service costs. Clarify export, handover and maintenance arrangements.

A performance commitment needs a defined outcome, baseline, measurement period, client responsibilities, exclusions and remedy. Establish those terms for the particular implementation. The phrase “performance guarantee” is not a substitute for a testable acceptance criterion and a written agreement.

Avoid creating a new dependency on the person who built the system. Name the operational owner, document recovery procedures and train an appropriate backup. CSM’s fractional team can work alongside your people to develop that capability, with responsibilities and support agreed in the scope.

What is the best first step?

Choose one recurring piece of work that matters commercially and repeatedly relies on the same person. Bring a recent example, the person doing the work and the leader who owns the outcome. Ask what another competent colleague would need to complete it safely.

That conversation gives us a practical basis for discovery, a prototype and a sensible first scope. You can explore how to map a business process for AI agents and how to build an AI Commercial Brain for more detail on the process and shared foundation.

Questions for your leadership team

What would stop if a key colleague were unavailable for a month?

Which decisions depend on memory, private files or personal access?

Can another authorised person find the evidence and complete the work?

Which records do we trust, and who owns their quality?

What should people approve, and how is that approval enforced?

Which improvement would create measurable capacity, protect margin or reduce risk?

Who will maintain the playbooks, connections and recovery procedures?

What would we need to see in a pilot before extending the scope?

Full video transcript

The supplied video script is reproduced below.

What would your business be worth if your best people left tomorrow?

In a lot of businesses, the knowledge lives in a few people’s heads. The systems don’t speak to each other. And success depends on a handful of individuals. That’s a risk to the business, and it’s a risk to what the business is worth.

We’re CSM. We help you build a business worth buying: more profitable, with stronger EBITDA, running more efficiently, on systems, with no dependency on any one person.

And we use AI to do it. But almost every business has access to the same AI now: Claude, ChatGPT, Gemini, Copilot. So the advantage isn’t the model. It’s what the system knows about your business: your data, your knowledge, your people’s judgement and the way you work. That’s why you can’t become an AI-first business until you become a data-first business.

We believe systems run the business, and people run the systems. AI works best inside good systems, run by good people. So AI does the gathering, the preparing and the follow-up, and your people bring the judgement, the relationships and the creativity your customers value. When the three work in harmony, the business needs less chasing and makes better decisions.

That’s our mission. We help businesses save time and money by improving how their systems speak to each other, and we help their leaders make better decisions. Saving time and money matters. Better decisions matter most: where to focus, what to protect and what to do next.

And that’s the outcome we work towards with every client. A business where the knowledge belongs to the company, reliable systems support growth, and success no longer depends on a few people. Whether you want to grow, acquire, attract investment or sell, a business that runs on systems is worth more.

So how do we do it? Every engagement follows the same seven steps, so you always know what happens next.

It starts with discovery. Before anything is built, we learn how your business wins and delivers work, and we agree with your leaders the decisions that matter most. Then we follow one real piece of work from the first request to the finished result, and we find where it gets stuck: where people search, retype, wait and check.

Every department is mapped, and every opportunity is scored on our priority scorecard. What’s the commercial impact? How much time would it give back? What stops if one key person leaves? What’s the risk of doing nothing for another year? And are we ready to act safely? A job can be quick and still carry serious risk, so we never let a total hide it.

The top priorities become your ninety-day roadmap, each with an owner, an agreed scope and a measure of success. We build a working prototype on your real examples, so you see it working before the full build. We build the full system and test it on the difficult cases. Your people pilot it on real work and improve it. We go live with support agreed, and we train your team on the jobs they actually do, not demos. Then it’s back to scope for the next priority. Every ninety days, we score again and set the next roadmap, building on what’s already live, so progress compounds.

Underneath every system is the same foundation. Your leaders set the priorities at the start, but we build from the data up, in six layers.

First, your data. Through CoreOS, we’ve already built API connections with over three thousand tools, so we connect the systems you already use into one trusted source of truth.

Second, an AI operating layer: one shared place where your people, your knowledge, AI and your workflows come together.

Third, your knowledge, skills and playbooks: what your business knows and how good work gets done, so AI follows your way of working instead of asking your team to explain everything again.

Fourth, the right AI model for each task, tested on real work, so your company knowledge stays useful as the models change.

Fifth, agents. This is where we build your AI workforce. Every agent gets a job description: its role, and exactly the outcome you want. And it’s an iterative process. We train each agent over time to work autonomously. At the start, an agent might take on fifty percent of a task. Then we improve that to sixty, to seventy, to eighty percent. It won’t solve everything on day one. But the whole premise is an AI workforce built around your people, working in harmony together.

And sixth, the layer that brings all of this together: commercial intelligence. Because every system in the business is connected, we’ve created an AI brain for the business. It highlights what’s in the green, what’s in the amber and what’s in the red. It shows you where the constraint is, and where your leaders need to support the business. It gives the business a voice. So if your business could talk to you, what would it say?

That’s the beating heart of everything we build, and it’s one of our biggest unfair advantages. Over the last five years, we’ve built commercial intelligence into our own business and into more than fifty others, and we’ve spent hundreds of thousands, with a full team, building it.

We’ve served over two hundred clients, tracked five hundred and eighty million euro in revenue from our campaigns, and earned more than a hundred client reviews.

And you can see it in the platforms we’ve built: an operating system for a construction company’s leadership team, a mission control that puts operational warnings and commercial exposure in one view, a finance operating system that prepares month-end, and Site-to-Quote, which turns a site visit into a quote.

But don’t take our word for it.

Behind every system is our own team. Our AI work is led by Kevin O’Brien, with fifteen years in AI, alongside our software developers and our automation and AI specialists.

Every system we build is completely custom, around how your business works. It’s yours for life. We put performance guarantees in place, so you know what it has to deliver. And the time it saves can pay back the investment within a few months.

So where do you start? Bring us one piece of work that slows your team down. That’s enough to start. Book a thirty-minute call, and we’ll show you a relevant example and what a sensible first scope would involve.

Reduce key-person dependency with CSM

Bring one process that repeatedly slows your team down. Start with a digital discovery workshop to map the work, agree the priorities and define a sensible first implementation.