Maybank Service Intelligence | Maybank Advisors
Assessment Book a review
Menu

The platform

Maybank Service Intelligence

Service intelligence for professional services firms.

Maybank Service Intelligence applies the operating model first developed inside Pyxl to other professional services firms. It connects recurring work, institutional knowledge, operating standards and management signals that normally live across project systems, documents, inboxes and individual memory.

What it is

An operating layer around the systems a firm already runs.

Rather than adding another system for teams to maintain, MSI works across the ones where client work already lives. It reads what is changing, assembles the context a person would otherwise rebuild by hand, and carries that context into the workflows the firm actually runs.

Agents run inside that layer. They are one component of MSI, configured last, once the context and the review standards are in place.

How the layer stacks
01 Intelligence layer What the firm asks of it
02 Connected context The systems where work already lives
03 Delivery signals What is changing across accounts
04 Workflows The steps a team actually runs
05 Configured agents Scoped per client, under review

Maybank platform · the order the layer is built in

What it does

Client and delivery signals

Reads across the systems where client work already lives, rather than asking teams to maintain another one.

Client-health concerns

Surfaces accounts that need attention before a status meeting establishes it.

Portfolio-level context

Assembles the view across accounts that normally takes someone senior an afternoon.

Account information on demand

Provides account context without a project manager reconstructing it by hand.

Client context in workflows

Carries approved client information into the work instead of re-establishing it each time.

Reporting, research, content and QA

Supports the delivery workflows where they fit the firm’s model, each with its own review gate.

Reduced coordination work

Removes manual project-management and account-coordination steps, which is where the operating leverage came from.

Configured agent workflows

Agents run inside the layer, scoped per client, under the firm’s own review standards.

What it is not

What MSI does not do

A firm carrying professional liability needs to know where a platform stops. These are the limits, and they are set deliberately.

Not a replacement for your systems

Asana, Slack, Drive and email stay where they are. MSI reads across them.

Not a CRM or PSA

It reads CRM context. It does not become the system of record.

Not a chatbot

Configured workflows against your delivery model, not a general-purpose assistant.

Not an autonomous workforce

Every workflow has a named owner and a review gate. Writes are proposed and confirmed.

Not a tool you switch on

Each implementation is configured around one firm’s delivery model, systems, methods and governance requirements.

Two layers

The work, and the operating system around the work.

The objective is not simply to make individual tasks faster. It is to reduce the friction required to produce, coordinate, review and manage the work the firm sells. Those are two different problems and they show up in two different places on the P&L.

Layer one

The work

  • Research and synthesis
  • Recurring drafting
  • Knowledge retrieval
  • Reusable firm methods
  • Recurring deliverables
  • Review preparation

Layer two

The operating system around the work

  • Briefing
  • Handoffs
  • Status reporting
  • Information retrieval
  • Recurring reporting
  • Version awareness
  • Management visibility
  • Institutional knowledge

The first layer can improve delivery economics. The second can create additional operating leverage.

Pyxl Intelligence was the first operating implementation of this model, built by Maybank inside Pyxl. From August 2025 through Q2 2026, Pyxl expanded gross margin by +1,300 bps and EBITDA margin +1,800 bps while the company changed its delivery and operating model alongside the technology.

Pyxl is one internal operating example. Results will vary by work mix, cost structure, adoption and commercial decisions.

Before the platform, the operating model

See why a tool alone does not produce firm economics.

The 18-minute briefing shows what changed inside Pyxl before Maybank Service Intelligence became the client-facing operating model—and where the evidence stops.

Watch the 18-minute briefing

Four chapters · Every figure cited · Not a software demo

The surfaces

Six screens from a configured implementation.

What each agent can do, what it is allowed to read, what runs without being asked, how it is gated, what it recorded, and what it cost to run.

Client names are illustrative and people are shown by role. These are representative MSI configurations: the operating model and controls Maybank configures for a client environment. Where a behavior is verified as running inside Pyxl today it is labeled as such below.

Agent library
Maybank Advisors

Service Intelligence

Home Agents

Workspace

Conversations Knowledge Base Briefs Reports Settings

Admin

Clients Prompts Scheduling Operations Usage Audit log

Agent library

Browse your firm's configured agents

Launch the ones that are live; the rest are configured in sequence. Each agent is scoped to a single client and follows your firm's own review standards.

All Strategy Content Paid Media Reporting New Business Clients Operations

Strategy

Strategy

Discovery, Briefing, and Campaign Planning

Expands a human-authored strategic insight brief into a full campaign plan; will not run without the brief attached.

Coming soon

Strategy

New Client Kickoff

Prepares kickoff agendas, intake questionnaires, proposed Asana structures, and onboarding plans from Drive, Asana, and AM-provided input.

Strategy

Competitive Intelligence

Broad competitive sweeps with a mandatory limitations section preceding any findings.

Coming soon

Content

Content

Blog and Long-Form Content

Drafts long-form content with a mandatory fact-check workflow before publication.

Content

Organic Social

Drafts organic social content batched two weeks at a time, with platform-specific voice references and a read-aloud test.

Content

Email Marketing

Drafts email marketing copy that requires a 15–30 minute editorial pass before any send workflow.

Coming soon
01  The agent library MSI workflows are designed with explicit refusal and escalation conditions so missing information or low-confidence cases can be routed to human judgment rather than guessed through. Agents are scoped to one client, and several are designed not to run without a human-authored brief attached, so the review standard is enforced by the system rather than requested of the team.
Settings
Maybank Advisors

Service Intelligence

Home Agents

Workspace

Conversations Knowledge Base Briefs Reports Settings

Admin

Clients Prompts Scheduling Operations Usage Audit log

Settings

VERIFIED LIVE PYXL CAPABILITY — connected accounts and client context mapping

Appearance, connections, chat defaults and notifications.

Connected accounts

Authorize Service Intelligence to read context from the tools your firm already uses.

Slack Not connected

Lets agents read the channels and threads you already have access to. We never post on your behalf.

Connect
CRM Not connected

Lets agents read CRM context — companies, contacts, deals, engagements — the same way you can. Read-only.

Connect
Google Drive Not connected

Lets agents read files and folders mapped to each client. Read-only. We never modify or share your files.

Connect
Asana Not connected

Lets agents read projects and tasks mapped to each client. Where an admin enables writes, agents propose task updates that you confirm before anything is written.

Connect

Chat defaults

Pre-select a client and agent so teams start inside the right context.

Default client

No default

Default agent

No default
02  Connected accounts Read-only by default and scoped per client, with no posting on anyone's behalf. Where writes are enabled, an agent proposes the change and a person confirms it.
Scheduling
Maybank Advisors

Service Intelligence

Home Agents

Workspace

Conversations Knowledge Base Briefs Reports Settings

Admin

Clients Prompts Scheduling Operations Usage Audit log

Scheduling

VERIFIED LIVE PYXL CAPABILITY — scheduled operations, enable/disable, manual and dry runs

Schedule background jobs on a daily, weekly, monthly, custom-cron or one-time cadence. The platform supports scheduled and recurring operational runs. These schedules are the single source of truth for when the jobs run.

4 schedules. The dispatcher runs every minute; changes take effect on the next tick.

New schedule

Weekly Client Health Check Report

All-clients agent batch run Client Health Check Enabled Last: success

At 04:00 AM, only on Monday · 0 4 * * 1 (America/New_York)

Target: All clients · Prompt: custom · Delivery: aggregate, directors

Run now Disable Edit Runs (5)

Slack signals — all clients

Slack signals Enabled Last: success

On the hour, every 6 hours · 0 */6 * * * (UTC)

Target: All clients

Run now Disable Edit Runs (10)

Operations User — daily report

Operations User Enabled Last: success

At 12:00 AM, Tuesday through Saturday · 0 0 * * 2–6 (America/New_York)

Target: All clients

Run now Disable Edit Runs (8)
03  The schedule dispatcher Jobs that run on a schedule without being asked. A weekly client-health check runs across every account and delivers one aggregated report to directors, work a project manager would otherwise assemble by hand.
Scheduled operations
Maybank Advisors

Service Intelligence

Home Agents

Workspace

Conversations Knowledge Base Briefs Reports Settings

Admin

Clients Prompts Scheduling Operations Usage Audit log

Scheduled operations

Controls the nightly project-management job: overdue-task notices, hours check-ins and the daily operations report. A scheduled, machine-identity job. No model call is involved.

Enabled Live send Save settings

Enabled

Disable

When disabled, the scheduled job skips every run — no notices, no task comments, no preview post.

Dry run

Switch to dry run

Dry run posts one preview manifest to the preview channel and sends nothing to individuals. Turn off only when you are ready for real notices and task comments.

Preview channel and test run

The preview channel receives the dry-run manifest. It is usually a private channel, so it may not appear in the picker unless you are a member — use the manual ID field in that case.

(not in your visible channel list)  CH-PREVIEW-01 Clear
Run preview nowRuns against saved settings — save first to preview your latest edits.

Team task digest

Weekly per-team digests, configured per schedule. This dry-run gate is separate from the nightly job above — pilot digests here before flipping them live, independent of whatever state the nightly job is in.

Dry runSwitch to live
04  The control surface Two independent dry-run gates and an explicit live-send state. Nothing reaches a person until an administrator turns the gate off, so a pilot can be run and reversed without consequence.
Admin audit log
Maybank Advisors

Service Intelligence

Home Agents

Workspace

Conversations Knowledge Base Briefs Reports Settings

Admin

Clients Prompts Scheduling Operations Usage Audit log

Admin audit log

ILLUSTRATIVE MSI CONTROL — audit and accountability model

The accountability model: which action was taken, by whom, and what it affected.

WhenActorActionDetails
8/26/2026, 2:45:49 PM Operations admin schedule.run.invoke
schedule:sch-4f7a
{ "runId": "run-8c21d0" }
8/26/2026, 2:45:43 PM Operations admin schedule.update
schedule:sch-4f7a
{
  "jobKind": "agent-batch-run",
  "clientId": null,
  "timezone": "America/New_York",
  "cadenceType": "WEEKLY",
  "cronExpression": "0 4 * * 1",
  "destinationType": "CHANNEL",
  "directorDeliveryEnabled": true,
  "perClientDeliveryEnabled": false
}
8/26/2026, 2:09:23 PM Operations admin schedule.enabled.change
schedule:sch-4f7a
{ "to": true, "from": false }
8/25/2026, 6:29:09 AM Delivery director schedule.run.invoke
schedule:sch-4f7a
{ "runId": "run-71b9ae" }
8/25/2026, 6:26:17 AM Workspace admin client.integration.set
client:cli-2b9d
{ "projectIds": [ "prj-1042", "prj-1088" ] }
05  The audit trail MSI is designed to make automated and human actions reviewable, with accountability controls appropriate to the workflow. The screen shows the accountability model the governance practice installs. Actors are shown by role.
Usage
Maybank Advisors

Service Intelligence

Home Agents

Workspace

Conversations Knowledge Base Briefs Reports Settings

Admin

Clients Prompts Scheduling Operations Usage Audit log

Token & cost

ILLUSTRATIVE MSI CONTROL — usage and cost accountability

MSI is designed to support visibility into usage and operating cost by workflow or client where the underlying systems provide reliable attribution. Figures shown are representative, not measured output.

Input: 8,857,745    Output: 484,420

Tokens per day

Stacked by token type.

Aug 23Aug 26Aug 29

Estimated cost per day

List-rate estimate.

Aug 23Aug 26Aug 29

Per-client usage

Operational totals only — no content.

Top clients

By turns, last 7 days.

Acme Water NewCo Retail Northwind Beverage Initech Pharmacy Globex Confections Contoso Group Umbrella Wellness Stark Home Wayne Studio Hooli Park
Client Turns Tokens
Acme Water7353,518
NewCo Retail5214,890
Northwind Beverage5295,802
06  The usage ledger Operational totals per client, no content. Cost is tracked as an internal reference figure rather than a billing rate, so the fee decision stays separate from the running cost.

The same operating pattern runs inside Pyxl as Pyxl Intelligence, which is the proof case rather than the product. See it in the Pyxl case study.

How it arrives

MSI was not built to be sold. It was built because Pyxl needed it, over 18 months of investment in the system, the processes and the prompt library, and it has been running there since. A client configures a system that already works rather than funding its invention, which is the difference between a project measured in months and one measured in years.

Configured during the engagement rather than licensed

MSI is scoped in the readiness assessment and configured during the build phase, against the firm's own delivery model, systems, methods and governance requirements. There is no seat count to buy and nothing to switch on.

The operating pattern was proved before it was offered. It runs inside Pyxl as Pyxl Intelligence, which is the proof case rather than the product.

“As Pyxl’s long-time Controller, I see the impact of AI through Pyxl Intelligence in the financial results, not just in anecdotal productivity gains. Over the past nine months, Pyxl has expanded gross margin by 13 percentage points and EBITDA margin by 18 percentage points. The gross-margin improvement reflects stronger delivery economics, and that gain flowed through to EBITDA. EBITDA expanded further as the company also generated additional operating leverage below gross profit. From a finance perspective, this is a meaningful improvement in the operating model, not simply a technology-efficiency story.”

Alban Howorth Controller, Pyxl and Founder, Howorth Accounting Pyxl Intelligence is the operating-intelligence system Maybank built inside Pyxl.
Book a walkthrough See the Pyxl proof →

Where it is scoped

In the readiness assessment. $4,500, two weeks, credited in full against a build engagement.

Where it is built

In the build phase. Configuration rather than construction where the delivery model fits, which shortens it considerably.

Who holds it afterwards

The firm. Prompt libraries, review standards and documentation sit in your own systems.

What it costs →