Groupe Conseil Lartis Inc.

AI  ·  Automation  ·  SMBs

AI for your
operations

Quickly identify where time is being lost, which repetitive tasks are worth automating, and where to start with practical, realistic solutions.

See our projects Contact us

20+years of hands-on experience
30+projects delivered
70–80%time savings on selected processes
Diagnostic — 2 minutes

Where to start?

Check the symptoms you recognize. Each symptom points you to the tool sheet that best fits the problem.

🧭
This guide is not meant to be read cover to cover. It is designed to be used from a concrete operational issue. Identify where time is being lost, then go directly to the relevant tool sheet.
"My employees spend time searching for information that already exists somewhere."
"I need to write the same emails, reports or summaries several times a week."
"Our decisions are often made without fresh data because reports take too long to produce."
"There are recurring data entry errors — copy-pasting between systems."
"My new hires always ask the same questions to the same experts."
"Our inboxes are overflowing — prioritizing and extracting information takes hours."
"Anomalies in our data go unnoticed until they cause a problem."
"Preparing a response to an RFP takes several days of work."
"When a key expert goes on vacation — or leaves — we lose critical knowledge."
"We manually check discrepancies between our ERP, CRM and spreadsheet tools."
Did you check one or more symptoms? Section B contains the corresponding tool sheet. Every week without action has a cost; Section A helps you estimate it.
A
The 5 operational loss zones

Where your time is being lost — and what it costs

In 20 years of SMB optimization, the same loss zones keep appearing. Identify yours. Estimate the real cost. Act on one zone at a time.

ZoneTypical symptomWhat AI can doEstimated cost / week
A1 — Scattered Information"Where is that file?" · "Who has the latest version?" · Searching in 4 different places.Assistant connected to your internal documents, with source-based answers.2–4 h/employee
A2 — Repetitive CommunicationSame emails rephrased · Reports that barely change · Manual meeting summaries.Assisted writing · Automatic extraction · Structured summary in 3 minutes.3–5 h/employee
A3 — Decisions made without timely dataReports too long · Data too old · Unreadable tables in meetings.Plain-language dashboard · Automatic anomaly detection · Summary in 30 seconds.Costly indecision
A4 — Avoidable Manual EntryCopy-paste between software · Dual ERP/CRM entry · Recurring transcription errors.Automatic data bridge · Anomaly detection · Reconciliation without human intervention.1 h+/day
A5 — Knowledge not yet capturedExpert answers the same questions · Knowledge leaves with the person · Onboarding redone each time.Interactive knowledge base · Internal conversational FAQ · Capturing and structuring operational knowledge.Structural risk
Lartis Field Rule: Start with the zone where the pain is most visible to your team, not the one that seems technically simpler. Adoption comes from the relief felt, not from the sophistication of the tool.
B
The 6 Tool Sheets

Each sheet = one problem solved

Symptom identified → Consult the sheet → Use the request template → Measure the result. Each sheet is standalone.

Sheet 01 · Zone A2
Email Triage & Extraction
📬
You recognize this problem if…

Your inbox is a constant stream where everything seems urgent. Finding key information in 50 messages takes 45 minutes of your morning.

AI approach and tools
  • Copilot M365 Triage and summary in Outlook
  • Claude Extract actions and deadlines
  • Make / Zapier Automatic classification
Request template (R.C.T.C.)

[R] You are a senior executive assistant.
[C] Here are 10 emails received today: [paste]
[T] Sort by priority. Extract: sender, request, required deadline, suggested action.
[C] 4-column table. If info missing: [MISSING].

Sheet 02 · Zone A2 & A3
Report & Summary Generation
📄
You recognize this problem if…

Turning a meeting or an Excel export into a readable report takes 1 to 3 hours. The report is often skimmed in 2 minutes.

AI approach and tools
  • Fathom / Otter Auto meeting summaries
  • Claude Structured visit report
  • Copilot Word/Excel synthesis
Request template (R.C.T.C.)

[R] You are a senior SMB optimization consultant.
[C] Here are my raw meeting notes: [paste]
[T] Generate: executive summary (3 sentences), decisions made, actions + owners + deadlines.
[C] Max 300 words. Professional tone. If info missing: [MISSING].

Sheet 03 · Zone A3 & A4
Data Anomaly Detection
🔎
You recognize this problem if…

Errors in your files, ERP or CRM go unnoticed for weeks. Their detection is manual, random, and too late.

AI approach and tools
  • Claude CSV/Excel file analysis
  • Power BI + Copilot Automatic alerts
  • Make Notification if threshold exceeded
Request template (R.C.T.C.)

[R] You are a data auditor for SMBs.
[C] Here is our data export: [paste table]
[T] Identify: duplicates, outliers, critical empty fields, logical inconsistencies.
[C] Table: Row | Problem | Severity | Suggestion.

Sheet 04 · Zone A1 & A5
Internal Assistant on Your Procedures
💡
You recognize this problem if…

The same questions keep coming up. Experts are consulted for things already documented — or that should be.

AI approach and tools
  • SharePoint + Copilot FAQ on your M365 files
  • Notion AI Interactive knowledge base
  • Custom Lartis solution Assistant connected to your internal documents
How it works

The AI is connected to your existing procedures (PDF, Word, SharePoint). It answers team questions while citing its source. It cannot invent what is not in your documents.

Sheet 05 · Zone A4
Data Reconciliation Between Systems
🔗
You recognize this problem if…

You have multiple software systems that don't talk to each other. Reconciling ERP, CRM and spreadsheets is done manually — with inevitable errors.

AI approach and tools
  • Claude Comparison of two exports
  • Make / n8n Automatic bridge between apps
  • Lartis Integration Zero re-entry, custom-built
Request template (R.C.T.C.)

[R] You are a data reconciliation expert.
[C] Export A (CRM): [paste] · Export B (ERP): [paste]
[T] List discrepancies, duplicates and missing records.
[C] Table: ID | Discrepancy | Missing source | Priority.

Sheet 06 · Zone A2
Assisted RFP Preparation
📋
You recognize this problem if…

Responding to a serious RFP mobilizes 2 to 3 people for several days. The first draft is often the most time-consuming part.

AI approach and tools
  • Claude RFP requirements analysis
  • Claude Structured response draft
  • Gamma Presentation in 30 seconds
Request template (R.C.T.C.)

[R] You are the bid director of a professional services SMB.
[C] RFP: [paste]. Our strengths: [paste our profile]
[T] Analyze the criteria, identify our arguments, generate a section-by-section draft.
[C] Formal tone. If info missing: [TO COMPLETE].

C
ARIA Method — Deploy Without Improvising

From idea to measurable result in 4 steps

Deploying AI without a method is like buying a tool without knowing what problem it solves. ARIA is the framework we have applied for 20 years to operations optimization — now adapted to AI.

A

Analysis

Map the real cost of the status quo. No assumptions — measurements. How many hours? How many errors?

R

Recommendation

Choose the right lever, not the most impressive one. A no-code tool that works beats a complex integration that drags on for 6 months.

I

Implementation

Start small. One task, one tool, one team. Measure from the first week. Adoption comes from the relief.

A

Adjustment

Optimize based on real data. What works: extend. What disappoints: understand before abandoning.

Field example — Food distribution client, 45 employees
Analysis3 employees spent 45 min/morning manually compiling inventory reports. Annual cost: ~$45,000 in non-productive time.
RecommendationAutomation via Make + Claude on existing ERP exports. No system overhaul. Budget: under $200/month.
ImplementationDeployed in 3 weeks. 2-hour training. The report now arrives at 7 AM in their inbox — structured, ready for action.
AdjustmentAfter 30 days: automatic alerts added for stockouts. The team itself identified the next automation.
D
Talking to Your AI Assistant

Writing effective AI requests — How to get useful results

AI is not just a search engine. It is a tool you give a clear assignment to. A vague request usually produces a vague result. A precise request produces something much easier to use.

The R.C.T.C. method — Four elements to include in every effective AI request: Role, Context, Task, Constraints.
R
Role
Who are you?

Give a precise role. "You are a consultant specializing in Quebec SMB optimization." The role shapes the tone, level and perspective of the entire response.

C
Context
What is the situation?

Describe your company, your sector, your client. The more precise the context, the less the AI invents. This is your insurance against errors.

T
Task
What needs to happen?

One clear action. "Write a 3-section report." Avoid multiple requests in a single prompt. Break it up if needed.

C
Constraints
What are the limits?

Format, length, forbidden tone. "Maximum 150 words. No technical jargon. If info missing: write [MISSING]."

Ready-to-use request templates by task type

WRITINGClient follow-up email
[R] You are a senior account manager at a professional services SMB.
[C] Our client [Sector] attended a meeting on [Date]. Topics covered: [summary].
[T] Write a follow-up email: thanks, recap of points discussed, clear next steps.
[C] Warm but professional tone. 3 paragraphs max. No jargon. End with an open question.
ANALYSISData synthesis
[R] You are a business analyst specializing in manufacturing / distribution / services SMBs.
[C] Here is our quarterly data: [paste export]. Our context: [brief description].
[T] Identify the 3 main trends, the 2 warning signals and formulate one actionable recommendation.
[C] If a number seems abnormal, say so explicitly. Do not invent any data. Cite your source.
EXTRACTIONMeeting notes → actions
[R] You are an executive assistant at a [sector] SMB.
[C] Here is the raw transcript of our meeting on [Date]: [paste].
[T] Extract: 1) Summary in 5 lines max. 2) Decisions made. 3) Actions = Who does What before When.
[C] Table for actions. If an owner or deadline is missing: [NOT DEFINED]. No invention.
E
Security & Confidentiality

Should AI be used for this task?

You do not need a long list of rules. Before using AI, ask two questions: are the data sensitive, and is the operational value high enough?

Low-sensitivity data
Highly sensitive data
Operational value →
High value
Go ahead
This is the sweet spot. Automate, test, measure. The value justifies the effort and the data poses no major risk.
High value⚠️
With precautions
Anonymize before transmitting. Use an approved tool (Copilot M365 if you're a Microsoft organization). Consult your advisor.
Low value🤷
Optional
Marginal gain. Do not prioritize. Start with a high-value case before tackling minor gains.
Low value🚫
Do not proceed
Double risk: exposed data for a marginal benefit. Refuse firmly. The rule applies to everyone — no exceptions.
← X axis: data sensitivity →  |  Y axis: operational value (high at top → low at bottom)
Law 25What Quebec's privacy law changes for your SMB
  • Any AI tool handling personal information about clients or employees is subject to Law 25.
  • You must document which data goes where — and for what purpose.
  • Copilot M365 is generally compliant: your data stays within your Microsoft tenant.
  • Third-party tools (Make, Zapier, Fathom…): verify the hosting location and sign a confidentiality agreement.
  • A register of personal data flows is required. Simple to create — indispensable in case of audit.
  • When in doubt about a specific case: consult your Privacy Protection Officer (PPO).

What repetitive task is frustrating your team the most right now?

That is often where you should begin. Not with a three-year AI strategy. Not with a transformation committee. With one task. One team. One test this week.

1
Identify the task (Diagnostic)
2
Time the current effort
3
Consult the tool sheet (Section B)
4
Run a first AI request
5
Measure. Decide. Move on.
Free AI Diagnostic — 30 minutes with Sylvain Laurence
📞 438-394-0394 info@lartis.com 🌐 lartis.com 📍 Laval, Quebec

Contact us

Have a project in mind? A question about automating your processes?
Write to us — we reply within 24 hours.

Contact

  • 438-394-0394
  • http://www.lartis.com