Skip to main content
HomeServicesApproachUse CasesAboutBlogContact
Change theme
Managed Services (MSP)AI Co-pilotIT / Managed Services

From doing to overseeing: Philippe gains time for his clients

At ProsperIT, Jadzia — the custom-built AI tool — captures client requests across every channel (email, text, tickets), makes sure they’re complete before opening a ticket, watches active tickets and flags what’s at risk of slipping. Nothing falls through the cracks.

ClientProsperIT

ProsperIT is a managed service provider (MSP) supporting its clients on Microsoft 365.

Process

Before vs After

Compare the current workflow with the AI-assisted process.

Before

Report received

A client flags a suspicious email

Figuring out how

How do you even analyze a header?

Drafting the reply

Manual wording, word by word

Formatting

Copy, format, double-check

Sending

In the right thread, to the right people

~45 min, variable quality

Depending on the technician’s experience

After

AIMulti-channel intake

Email, text, tickets: everything converges on Jadzia

AIClarifying questions

She completes the request before opening a ticket

AIAnalysis and action

Diagnosis, reply drafted, PDF generated

HumanSupervision

Philippe approves sensitive actions

SystemComplete ticket created

With all relevant information · Slack trace

AIWatch and reminders

Jadzia flags at-risk tickets to Philippe

Nothing falls through the cracks

Every client request carried to completion

Quality Control

Smart Validation

The agent evaluates its own confidence and adapts its behavior.

Autonomous

Jadzia acts

Informational replies, reports, scheduling proposals. The ticket and Slack trace are created automatically.

Quick review

Philippe takes a look

Direct reply to a client, sending a PDF. Philippe checks before it goes out.

Approval

Philippe decides

Security incident response, access or account change. Explicit approval required.

Day to Day

A Colleague, Not Software

Here's what a workday with the agent looks like.

1 / 5 — Morning watch: reminder to Philippe

Morning watch: reminder to Philippe

Today 8:00

Good morning Philippe! Here’s the morning watch on active tickets:

🟢 11 on track

🟡 3 to watch (no update in 48 h)

🔴 2 at risk of missing the deadline:

• PROSP-940 · Mailbox migration: waiting on the client for 4 days

• PROSP-952 · License renewal: due tomorrow, approval missing


Should I follow up with the clients on the 🔴 ones?

JadzIA
Yes, follow up on both. And escalate PROSP-952 if there’s no answer by noon.
On it:

✅ Follow-ups sent (text + email) on PROSP-940 and PROSP-952

⏰ Noon alert set on PROSP-952, I’ll escalate to you if nothing moves


I’ll give you an update this afternoon.

JadzIA · ProsperIT team

Read at 8:07

Testimonial

“Jadzia consolidates requests and tickets, and makes sure nothing falls through the cracks. I save about ten hours a week, and since everything is documented automatically, I finally have real peace of mind.”

Philippe Daoust

President, ProsperIT

Results

Measured Impact

The numbers speak for themselves.

Before

45 min / incident

After

~5 min

to analyze and respond to a security report

Before

Variable

After

Consistent

service quality, regardless of the technician’s experience

Before

Manual

After

Automatic

ticket + Slack trace created for every action

Before

Doing

After

Overseeing

the technician handles more clients, with less friction

Under the Hood

Technical Architecture

The technical layers of the agent, from input to output.

1

Multi-channel intake

Email, text, tickets · Microsoft Graph (calendars, tenants), Defender · multi-tenant

2

Enrichment

Clarifying questions to complete every request before the ticket

3

Analysis

Header reading, diagnosis, generation of branded reports and PDFs

4

Supervised action

Reply in the right thread, signed on the technician’s behalf · approval of sensitive actions

5

Watch and traceability

Active-ticket monitoring · risk-level reminder to Philippe · Slack trace

Have a Similar Project?

We can build an AI agent tailored to your processes.