AI implementation
AI implementation, from advice to a running system.
We are consultants who operate. Once the advice is given, we build it with you: the workflow redesigned, the AI wired into ERP, commerce and CRM, governed from day one, rolled out in stages, and handed over so it runs without us.
Advice is the easy half.
Good AI advice is everywhere. A running system is the part we deliver: scoped, built, integrated with the systems that hold the real work, and live in the hands of the people who do that work every day. The industry numbers say this is where the money goes to die: MIT NANDA found 95% of enterprise generative-AI pilots deliver no measurable P&L impact, and Gartner forecasts 30% of generative-AI projects abandoned after proof of concept.
The demo is the easy part. What separates the pilots that die from the systems that pay is implementation: the workflow, the wiring, the governance and the rollout. We wrote up the whole pattern in From pilot to production, our guide on why most AI dies in the demo.
What implementation covers
- Workflow redesign. The process is rebuilt around the AI before any code is written, so the system has real work to do.
- Integration with the systems that hold the work. ERP, commerce, CRM, warehouse and store systems: the model connects to where orders, stock and customers actually live.
- Data plumbing. The pipelines, permissions and quality checks that feed the system, built for the data you have.
- Governance and audit trail. Controls, accountability and a trail a reviewer can walk, designed in from the first workflow so the system can pass review and stay in production.
- Staged rollout with fallback. Live in stages, with a way back at every stage.
- Handover. Documentation, training and an operating rhythm, so it runs without us. That is the finish line we aim at.
What integration services means in practice
Connecting an agent to a demo dataset takes an afternoon. Connecting it to a legacy ERP with fields named by someone who left years ago, three naming conventions and edge cases nobody documented is the actual job, and it is the part we plan for from the first session. We map the edge cases up front, decide build versus buy per component (our build vs buy guide has the decision table), and wire models and agents to your ERP, commerce and CRM realities rather than to a sandbox.
This is the build half of AI in production, our program for AI that is built into the workflow, run end to end, and governed so it scales past the pilot.
Who does the work
Senior operators lead every engagement, with a dedicated build team behind them and the 100+ AI agents we run to build faster. We ship from inside the discipline we sell. The same team put a price-tag platform live in 150+ stores across 10 countries, so we know what it takes for a system to hold up in the places where work is messy: stores, warehouses and head offices at the end of a long integration chain.
Questions, answered straight.
What is an AI implementation consultant?
An AI implementation consultant takes AI from a recommendation to a system running inside your business. Where an advisory engagement ends with a report, implementation covers workflow redesign, integration with your ERP, commerce and CRM systems, data plumbing, governance, staged rollout and handover. We work as consultants who operate: once the advice is given, we build it with you.
How much does an AI consultant cost?
It depends on the shape of the engagement, and the honest answer is a scoping conversation. The common models are a fixed-scope diagnostic, a fixed-price build with defined deliverables, and a monthly engagement for longer programs. The cost drivers are integration depth (how many systems the AI must touch), the condition of your data, governance requirements, and how much of the rollout you want run for you rather than advised. We price after a working session, once the real scope is visible.
What do AI integration services include?
Connecting models and agents to the systems where the work actually lives: ERP, commerce platforms, CRM, warehouse and store systems. That includes the APIs and data pipelines, the legacy fields and edge cases that never made the documentation, permissions and an audit trail, and fallback behavior for when something upstream goes down. Integration is where most of the effort goes in a serious deployment, and where most pilots quietly stall.
Do we need a consultant, or should we build or buy this ourselves?
Sometimes neither, and we will say so. Buy by default where the capability is a commodity, build where it touches your edge, and bring in outside help when the integration crosses more systems than any one internal team owns. Our build vs buy guide has the decision table we use. If the right answer is buy and configure, we tell you that in the working session and save you the engagement.
See what implementation would look like in your business.
A working session with the senior people who would do the work. If a scoped first step helps, the FOCAL Diagnostic ends in a written read you can take to your board.