Strategy
AI strategy and readiness
A structured review of your operations that ranks opportunities by value, risk and effort, and names the ones to leave alone.
Applied AI advisory and implementation
AI put to work, with the controls a serious business requires.
Frank Potvin Group advises owners and leadership teams on where artificial intelligence earns its place in their operations. We build the systems that do, and we remain accountable for them after launch.
01 / Where clients start
Most engagements begin with one of four situations.
- i.
Skilled staff spend hours each week moving information between email, spreadsheets and the systems the business runs on.
- ii.
Leadership has set aside a budget for AI but has no agreed view of where it should be spent first.
- iii.
Pilots were run and tools were licensed, and nobody can say with confidence what they returned.
- iv.
A critical process depends on one person, and it slows down whenever that person is away.
-
Step 1
Intake
Email, forms, documents and the systems you already run.
-
Step 2
Rules first
Deterministic checks decide every case they can.
-
Step 3
Bounded model
A model handles only what rules cannot, with a fixed set of permitted actions.
-
Step 4
Human review
Low-confidence or high-value cases go to a named person.
-
Step 5
Record
Each decision is logged for audit and for improvement.
Escalation
At any step, a case below the confidence threshold stops and is routed to its owner with the source material and the one question that needs an answer.
02 / Practice areas
Six areas of work, each with a defined outcome.
Clients often start with an assessment and move into one or two of the other areas. Each can also be engaged on its own.
Implementation
Workflow automation
The repetitive path between your existing systems, built once and handed over with documentation your team can maintain.
Implementation
Assistants and agents
Assistants with a defined role, defined permissions and a defined escalation path, deployed inside your own environment.
Implementation
Document and knowledge retrieval
Search and question answering over your own records, with every answer linked to the source it came from.
Operations
Governance and managed operation
Monitoring, cost reporting and periodic review, so that a system which was right at launch stays right as the business changes.
Enablement
Training for leadership and staff
Practical sessions on productive and safe use of AI tools, built around your own documents and policies.
03 / Approach
Four phases, each with a decision point.
You decide at the end of every phase whether to continue. Nothing moves to the next phase on assumption.
-
01
Assess
2 to 3 weeks
Interviews, process walk-throughs and a review of your data and systems.
Output A ranked opportunity register and a written recommendation.
-
02
Design
1 to 2 weeks
Scope, controls, success measures and a running-cost estimate based on your own volumes.
Output A design you approve before any build begins.
-
03
Build and validate
4 to 8 weeks
Delivered in stages and tested against real cases, with your staff reviewing results before anything runs unattended.
Output A working system, measured against the agreed baseline.
-
04
Operate
Ongoing
Hand-over to your team, or ongoing operation by us under a monthly agreement.
Output Regular reporting on accuracy, volume and running cost.
04 / Engagements
Commercial terms that are clear before work begins.
Every engagement is quoted in writing. Fees for build work are set after the assessment, so that the figure reflects the real scope rather than an estimate made in a first meeting.
Fixed fee
Readiness assessment
The usual first step. Two to three weeks, ending in a written recommendation you can act on with or without us.
Fixed fee per phase
Implementation
Design and build of a defined system, with acceptance criteria agreed in advance.
Monthly
Managed operation
Monitoring, maintenance, cost reporting and a quarterly review of the systems we have built.
Monthly
Advisory retainer
Senior access for leadership teams that want an independent view on AI vendors, proposals and policy.
05 / Principles
The standards every engagement is held to.
- Independent advice
- We do not resell software. A recommendation is never shaped by a licence margin or a vendor relationship.
- A process is documented before it is automated
- If a process exists only in one person's head, we write it down first. Automating an unclear process only makes the confusion faster.
- A named owner and an off switch
- Every system has a person accountable for it, a defined escalation path and a fallback the business can operate on without it.
- Running costs stated in advance
- You see the expected monthly cost, measured against your own volumes, before you commit to a build.
- Your data stays under your control
- We work inside your accounts and under your policies wherever possible, and access ends when the engagement does.
06 / Questions
Questions clients ask early.
How do you handle our data?
Access is limited to what the engagement requires, granted through your own accounts where possible, and recorded. At the end of the engagement it is removed. Where a third-party model is used, we select services and settings that do not train on your data.
Do we need to replace our existing systems?
Rarely. Most of the work connects to what you already use: email, document storage, accounting, CRM and line-of-business software. Replacing a core system is a separate decision, and we will say so if we think it is warranted.
What happens if something fails outside business hours?
Systems are designed to stop and escalate rather than guess. Under a managed operation agreement we monitor them and respond within the hours set out in that agreement.
Can our own staff maintain what you build?
Yes. Hand-over includes documentation written for your team, a walk-through, and full access to source code and configuration. Ongoing support from us is optional.
How do we know whether it worked?
Success measures are agreed during design, and a baseline is taken before the build starts. After launch we report against those same measures.
Contact
Start with a conversation.
A first call takes thirty minutes. Tell us which work takes too long, and we will tell you plainly whether AI is the right answer.
Email the firm
We reply within one business day.