Prepare to transform your enterprise with AI
Five things that decide whether AI changes how a large company works.
For thirty years, large companies have bought software the same way. A team names a problem, picks a tool, runs a project and goes live. A CRM for sales, a ticketing tool for support, a payroll system for HR. Each tool sits inside one department and helps the people there keep records and pass work along.
AI agents do the work itself. They answer the customer, check the policy, change the booking and chase the missing document. That work rarely stays inside one department. A refund request touches customer service, finance and operations. A network fault touches the call center, the field teams and billing.
The companies that get the most from AI plan for it the way they would plan a reorganization, with an order of steps, named owners and results the board can see. Companies that add a chatbot here and an assistant there end up where they started, with a few more tools.
In our work with enterprises across industries, five things decide which way it goes.
01Choose the first problem carefully
The first project sets the pace for everything after it. Pick one that matters to the business, that can go live soon, and that leaves something the next projects can reuse: access to the core systems, clean customer data, a way to review what agents do. Then the second and third projects start from that base instead of from zero.
Before you start, write down where you want the company to be in two or three years. The order of projects should lead there.
02Follow the work across departments
Trace the path of one real request, from the customer's first message to the last update in your systems. Automate that whole path. If an agent answers a billing question and then passes the customer to three teams, little has changed. If it answers, corrects the bill and closes the case, a whole piece of work is gone.
This usually means several agents working together and with people: one takes the call, one checks the account, one makes the change, and a person approves what needs a person.
03Build the skills inside your company
An outside partner can help you move fast at the start. Over time, the knowledge of how your agents run your business should sit with your own people. Name an owner in each business line. Train teams to read what agents do, correct them and design the next ones. Ask every partner, including us, how and when they will hand over.
04Keep your options open
AI models change every few months. Prices fall, better models arrive, data rules get stricter. Build so you can switch model, vendor or hosting without starting again. Keep your data, instructions and workflows in a form you own. Decide who can approve what an agent does, and keep a record of every action it takes.
05Measure what the business gets done
Cost per call is the easiest number to track, and it misses most of the value. Also count what the company can do now that it couldn't before: requests handled at night and on weekends, cases closed on first contact, sales made during service conversations, backlogs cleared. Report these next to revenue and cost, so the board sees AI in the business results.
Where to start
Pick one process that matters and decide where it should lead. Put it live, measure it, train your teams to run it, and use what you built to go faster on the next one.
Shapely works this way. Our engineers sit with your teams, start from the problem you bring, and stay until your people can run what we built together.
