AI adoption
Get AI working inside your business.
Buying the tool was never the hard part. Deciding how the business uses it is.
Some companies bought Copilot or Gemini and nothing changed. Some have not started. Either way, four business decisions have not been made: which processes to rebuild around the tool, who owns each one, what the rules are for using AI with company data, and what each process is measured against. ISC works through those four decisions with you.
Built for companies between $50M and $500M in revenue. Available to larger companies as part of a program.
The problem
IT did their job. The business has not made its decisions.
Most mid-market companies have already bought an AI tool. Ask around and everyone says they use it. Look closer and it is writing emails, rewriting paragraphs, and producing proposals a manager has to redo. The usage report looks fine. The work costs the same as before and the quality has not moved.
If someone asks whether you have deployed AI, the answer is yes. That box is checked. Checking it was never the point. The point is fewer hours per proposal, faster quotes, fewer errors in the back office, and a number on the P&L you can point to.
Companies that have not bought anything are in a similar position for a different reason. The vendors are calling, every demo looks good, and the company has not decided which process should go first, what the governance rules should be, or what a good result looks like in numbers. So nothing gets bought, or the wrong thing does.
In both cases the company has not made the business decisions that determine whether AI produces a return: which processes change, who owns each one, what the governance rules are, and what each process is measured against.
What we find
What we find in companies this size
Meeting notes and faster email are the gains most companies get from AI, and they are small. What pays for an assessment is finding one process problem like the five below. The rebuild does not depend on whether you already own a tool or have not bought one yet.
Finding 1
Quotes that go out second.
A distributor or a manufacturer with a quote desk. A request comes in by email. An estimator checks stock, checks the price book, checks the exceptions spreadsheet for that customer, and writes the quote. Two days, sometimes three. If the company owns an AI tool, it is being used to improve the wording of the email. The rebuild: the tool reads the request, pulls stock and price, drafts the quote, and routes the exceptions to an estimator for approval. Same day. On a $120M book, the difference between quoting first and quoting second shows up in win rate, and one point of win rate is several times the fee.
Finding 2
Scope changes that never get billed.
A contractor. A superintendent tells the office in a Teams message that the owner wants the wall moved. The work gets done. The change order gets written up three weeks later, or not at all. The rebuild: every week, the tool reads the project channels and email and produces a list of conversations that look like scope changes with no change order attached. A project manager confirms or dismisses each one. One recovered change order pays for the assessment.
Finding 3
Contracts that renew themselves.
Back office. Vendor agreements live in SharePoint, in email attachments, and in a drawer. Auto-renewals fire. No one renegotiated because no one knew the date. The rebuild: the tool reads every agreement it can reach, builds the renewal calendar with the notice windows, and the CFO gets a list 90 days out. In a company with two or three hundred vendors, the first pass finds money.
Finding 4
The person who is retiring.
Somewhere in your company there is one person who knows why Customer X gets 12% off, which jobs never get bid, and which vendor invoice is always wrong. That person is 63 and wants to retire soon. When they leave, that knowledge leaves with them, and no AI tool can answer a question that was never written down. The rebuild: document the rules as rules, with that person, while they are still there, and store them where the tool and their replacement can both use them. CEOs tell us this is the finding they were not expecting.
Finding 5
Proposals a manager has to redo.
The sales team writes proposals with AI in an afternoon. The VP spends 90 minutes fixing each one because they read like every other proposal the tool has ever written. The rebuild: a proposal process with a real intake (what the customer said, what we are pricing, what we are not), your last three winning proposals as the reference, and a review step before it reaches the VP. Review drops to 20 minutes and the proposals sound like your company.
The numbers above are illustrative for companies of this size. In an assessment, yours are measured against a baseline before anything changes.
One company, one process
A $120M distributor. 280 employees. 150 Copilot seats.
Copilot is the example because it is the tool most companies this size own. The steps do not change on Gemini, Agentforce, or ChatGPT Enterprise.
Today
Monday, 8:40 AM. Forty quote requests in a shared inbox. Three estimators. Each one pulls stock from the ERP, prices from the book, and exceptions from a spreadsheet only two people trust. Copilot tidies the email. The last quote in the batch goes out Wednesday afternoon. Two customers already bought elsewhere. No one knows which two. The Copilot usage report says 78% of seats are active, and the CFO approves the renewal.
After the first rebuild
Monday, 8:40 AM. Same forty requests. Copilot has read each one, matched it to stock and price, drafted the quote, and routed the twelve with exceptions to an estimator. The estimator approves or changes them. Thirty-eight quotes are out by noon. The two that need a phone call get a phone call. The estimators spend the afternoon on the accounts that make up 60% of margin. Every quote carries a timestamp and an outcome, so by Friday you know your win rate by turnaround time. That is a number you have never had.
What changed
- Quote turnaround: two to three days, to same day.
- Estimator time per quote: 45 minutes, to 10.
- Win rate: measured for the first time, against a baseline, so the next pricing decision is made with a number.
- The rules: documented, used by the tool today and by the next estimator you hire.
None of this required custom software. It required three decisions: which process, whose rules, and what gets measured. The assessment produces those decisions.
The AI Adoption Framework
Define. Analyze. Execute. Applied to AI.
Microsoft publishes an adoption framework for AI. So do the other vendors. They are good, and they are written from the technology side: licenses, tenants, connectors, security settings. This one starts from the process. Which work changes, whose rules apply, and what number tells you it worked. The technology decisions come after those and are easier once they are made.
ISC runs these three phases on every engagement. On an assessment they take four weeks. On a program they take months. The questions do not change.
Phase 1
Define
Decide what you are doing and why.
Every engagement begins with the business problem, stated by the executives who own it. Not the technology initiative. A CEO who says 'we need to do something with AI' has not defined a problem yet. A CEO who says 'our quotes go out two days after the competition' has defined the problem.
Step 01
Define the problem.
What costs you money, time, or customers today? State it in one sentence with a number in it.
Step 02
Define what good looks like.
For each process, one number, measured before anything changes. Hours per quote. Days to close a change order. Rework rate. Adoption rate is not on the list.
Step 03
Select the processes.
Against four conditions: the work recurs, it is costly when it goes wrong, it can be measured, and a named process owner exists. Three processes, ranked by volume and effort.
Phase 2
Analyze
Find out what is in the way.
This is where most AI projects fail before they start. The tool is deployed, the process is selected, and then it turns out the rules live in one person's head, the tool cannot read the spreadsheet everyone uses, and the company has not decided what the tool is allowed to see. Analyze finds those problems before the rebuild, not during it.
Step 04
Tools and licenses.
What you own, what it is producing, and what to stop paying for. Or, if you have not bought anything, which tools fit the processes you selected and which do not. Buy, build, or extend what you already have. Most mid-market companies should extend.
Step 05
Data readiness.
An AI tool answers from what it can reach. If the price exceptions live in a spreadsheet two people trust, the tool cannot use them and will produce an answer from whatever it can read instead. This step finds where the rules, records, and reference documents actually live, whether the tool can read them, and whether they are current enough to trust. The technical term for connecting a tool to your own documents is retrieval-augmented generation. In practice it means putting the rules where the tool can find them.
Step 06
Security and sensitive information.
The tool sees what the user can see. If a warehouse supervisor can open the payroll folder today, the AI tool can read it for them tomorrow. This step reviews permissions, decides what stays out of reach, and sets the rules for what leaves the company. Customer data, pricing, contracts, and employee records each get an answer.
Step 07
Governance.
What your people can and cannot do with AI and company data, written for your company. Who approves a new use. Who owns each rebuilt process. Whether you need an AI council or whether one meeting a quarter is enough. What responsible use means in practice: the tool drafts, a named role approves, and that role is accountable for the result.
Phase 3
Execute
Rebuild one process and measure it.
Execution is where transformations fail, and it is usually because the work was handed to IT or to a vendor and the business walked away. Execute is done with the process owner and the SMEs, in your business, against the baseline from the Define phase.
Step 08
Baseline and rebuild.
Measure the process as it runs today. Then redesign it around the tool: the tool handles the reading, matching, and drafting, and a named role approves the exceptions. One process, in production, before starting the next.
Step 09
Own, measure, extend.
The process owner runs the redesigned process. The number from Step 02 is checked every week. When it holds, the next process starts. Companies that want more extend from there, one process at a time, with a champion in each department who has already done it once.
How we work
Fixed price. Fixed scope.
The assessment identifies where AI pays off in your business. The Rebuild puts the first process into production. Both are fixed price.
1. Assess
AI Operating Model Assessment
Four weeks. $18,500, fixed. No hourly billing. No change orders.
At the end you know where the money is, what needs to change first, who does it, and what it will be measured against.
You get
- A tool and license audit. If you own the tools: which seats are active, what they are producing, and what you can stop paying for. If you do not: which tools fit your business, which to skip, and what they will cost. ISC does not resell software and takes no commissions from any vendor.
- Three processes to rebuild first. Selected from your business against four conditions: the work recurs, it is costly when it goes wrong, it can be measured, and a named process owner exists. Each comes with the hours and dollars at stake and a baseline measured before anything changes.
- Governance rules. What your people can and cannot do with AI and company data. Written for your company, not copied from a template.
- A 90-day plan. Who owns each change, what order it happens in, and the number each one is judged on.
- A readout with your leadership team. If the evidence says do not proceed, the readout says so.
How it works
We start with one meeting with you and your leadership team to agree on what a good result looks like, stated as a business problem and not as a technology initiative. We pull the license data, walk each process with the process owner and the SMEs, and measure the current state. We end with the readout. Your time is three to four hours over the four weeks.
One thing we ask for up front: the names of the two or three people who will work with us, and your agreement to protect their time. Borrowed time gets pulled the week it matters.
2. Rebuild
Rebuild: one process, in production
Up to four months. Fixed price, based on the assessment readout.
The Rebuild takes one of the three processes from the assessment readout and puts it into production. We document the business rules with the process owner and the SMEs. We complete the data readiness work the assessment identified, so the tool can read the records, price lists, contracts, or reference documents the process depends on. We redesign the process around the tool, with exception handling and approval points assigned to named roles. We train the process owner and the SMEs, run the process in production for a measured period, and report results against the assessment baseline.
You get
- The redesigned process running in production.
- Documented business rules, stored where the tool and new staff can use them.
- The process owner and SMEs trained on the redesigned process.
- Results measured against the assessment baseline: hours, cycle time, error rate, and dollars.
- Any build the process requires, such as a Copilot Studio agent or a system integration, scoped into the price.
- A recommendation on which process to rebuild next.
Pricing
The price is fixed and based on the assessment readout: process volume, the data readiness work required, and any build work involved. Additional processes are scoped one at a time.
The assessment is complete on its own. The Rebuild is the next step for companies that want the first process in production.
Larger companies get this work inside a transformation or platform program. The case studies show what that looks like at scale.
Scope
What the assessment covers, and what needs its own engagement
The assessment is built for
- Companies between $50M and $500M in revenue.
- Companies that bought an AI tool and concluded it did not change much. That is a normal result and a good starting point.
- Companies that have not bought anything yet and want the decision made before they spend.
- Companies without a full-time AI executive and no plan to hire one.
- Leadership teams that will name the people who work on this and protect their time.
Needs its own engagement
- Custom agents and integrations. We scope them in the assessment and build them in a Rebuild.
- A rollout across the whole company. That is a transformation program, and ISC runs those.
- An internal platform team that wants a plan it can execute itself. The assessment gives them one.
Credentials
Certified on the platform most companies already own. Independent of all of them.
Rob Anonen, ISC's founder, holds two Microsoft certifications in AI transformation. ISC does not resell software, takes no commissions from any vendor, and runs the engagement on whichever platform you own.
Microsoft Certified: AI Business Professional (AB-730)
Microsoft Certified: AI Transformation Leader (AB-731)
Microsoft 365 Copilot and Copilot Studio, Google Workspace with Gemini, Salesforce Agentforce, ChatGPT Enterprise, and Anthropic Claude.
Next step
A 30-minute call. Bring one document.
Bring one proposal, quote, or report your team produced with AI last month. If you have not started, bring the process that costs you the most when it goes wrong. We read it together, and by the end of the call you know whether the assessment fits your company.