Agents · Workflows · Intelligent dashboards

Agents for judgment. Workflows for certainty.

One intelligent dashboard where your team sees both, and acts.

You own the what; I own the how, and I ship your changes in days, not weeks.

100%
deterministic where the rule is known
2–4 wk
from idea to something running in production
1
surface your team actually works in

The gap

Not everything should be an agent.

The last two years framed every problem as an agent. But most enterprise work isn't ambiguous. It's a known rule, run a thousand times a day. Wrapping a model around it adds cost, latency and a failure mode you didn't have before.

The work that does need judgment is real, and it's where agents earn their place. The craft is knowing which is which, and giving people one place to see and act on both.

Determinism first

If the rule is known, encode it. Predictable cost, predictable output, a clean audit trail.

Judgment where it pays

Agents for the ambiguous middle: reading, weighing, drafting, deciding what to escalate.

One surface

People shouldn't chase five tools. One dashboard that shows the state of the business and lets them act on it.

Governed throughout

Set the rules once. Observe every run. Roll back anything.

Sound familiar?

You tried. It half-worked. Then it stalled.

Maybe you connected ChatGPT or Claude to your mailbox and it was impressive for a week. Maybe you tried something bigger and it quietly died. Or maybe you look at all of it and genuinely don't know where to start.

All three are the normal place to be. None of them mean the tools are bad, or that you did it wrong. Here is where people actually get stuck.

01

You don't know what to automate first

Everything looks automatable, nothing feels safe to start with, so nothing ships. The hardest part isn't the technology, it's choosing.

We start where there is volume, a clear trigger and a willing owner.

02

Brilliant in the demo, then Monday happened

The first test is impressive. Then the exceptions arrive, and the exceptions turn out to be most of the job.

Rules cover the known cases. The agent only takes what is left.

03

You can't trust it with anything that costs money

Same question on Tuesday, different answer on Thursday. Fine for drafting a reply. Not fine for a price, a date or a commitment.

Deterministic where it matters. A number never comes from a model.

04

It only works when you're at the keyboard

An assistant waits to be asked. The request that costs you lands at 22:00 on a Sunday, and it's still sitting there Monday morning.

Triggered by the event, not by you opening a tab.

05

The emails it reads can tell it what to do

Everything in an incoming message is input. Someone can write instructions inside an email, and an assistant with access to your mailbox may simply follow them.

Scoped access, and nothing consequential without a check.

06

Nobody can see what it did

When it gets something wrong you can't explain it to the customer, can't find out why, and can't undo it.

Every run logged, every action reversible.

The sentence I hear most often is "I'm a bit worried the AI does something stupid." That is exactly the right worry, and it's the one nobody selling AI wants to talk about. The answer isn't a smarter model. It's deciding in advance what it is allowed to do on its own.

Tell me where you're stuck

The platform

Three layers. The top one is the only one your team has to think about.

Workflows and agents are the building blocks. Compose them and you get an intelligent dashboard, the place where your team sees what's happening and triggers what happens next. You tell me what you need. I design it, build it, host it, and turn your changes around in days.

Layer 1 · Certainty

Deterministic workflows

  • Triggered by an event, not a prompt
  • Connected to the systems you already run
  • Validation, retries, and a full audit trail
  • Same input, same output, every time
Layer 2 · Judgment

AI agents

  • Grounded in your context and your policies
  • Tool-using: they read, act, and hand back
  • Evaluated before they ship, monitored after
  • Human in the loop exactly where it matters
Layer 3 · Control

Intelligent dashboards

  • A composition of the workflows and agents below
  • See the state of the business in real time
  • Act from the same screen: trigger a run, ask an agent, update a record
  • Your system of record, reachable by the whole team from anywhere
  • Per-user permissions: who sees what, and who can trigger what
  • Changes to what's live ship in days; new integrations are quoted
  • Fully managed: nothing for you to install, host or patch

What I ship

One agent. The workflows behind it. The dashboard to pilot it.

Two agents cover most of the ground: one for sales, one for customer care. Each ships with its own deterministic workflows underneath and an intelligent dashboard to steer it: not three separate projects, one working system.

Sales & Marketing

AI sales agent

  • Agent
    Qualifies, researches, follows up
    Knows your offers, your pricing rules and your tone of voice
  • Workflows behind it
    Routing, enrichment, CRM sync
    Deterministic: same input, same result, every time
  • Dashboard
    Pipeline cockpit
    Pilot the agent, read every conversation, step in when it matters

Customer Care

AI customer care agent

  • Agent
    Resolves, grounded in your knowledge base
    Answers in your customer's language, escalates the moment it's unsure
  • Workflows behind it
    Triage, routing, SLA timers, escalation paths
    The rules never drift, and every step is logged
  • Dashboard
    Service desk
    Live queue health, and take over any conversation in one click

Operations & Finance

Workflow-led, agent-assisted

  • Workflows
    Invoice processing, reconciliation, onboarding
    Extract, validate, post, with a full audit trail
  • Agent
    Exception handler
    Picks up only what the rules couldn't close on their own
  • Dashboard
    Ops control room
    Exceptions surfaced, reruns triggered in place

The platform in production

A European telco took a customer-onboarding agent live in hours, and across its markets in about four weeks.

That's the platform your dashboard runs on, already proven at telco scale. I bring it to your business, under your brand, with the workflows and the dashboard built around it.

Hours
to the first agent in production
~4 wk
to rollout across markets

How I work

Automate first. Add judgment only where the rule runs out.

01

Frame

Pick the work with volume and a clear trigger. Decide together what's a rule and what needs judgment.

02

Automate

Ship the deterministic path first, in 2–4 weeks. It pays for itself before any model is involved.

03

Add judgment

Introduce agents only where the rule runs out. Evaluated before they ship, monitored after.

04

Compose

Assemble the pieces into a dashboard your team runs the business from. The next one is faster than the last.

Tom Guisgand, founder of Guiz Tech Consulting

Who you'll be working with

Tom Guisgand

Fifteen years in customer experience design, the last few spent putting AI agents into production at consumer scale inside a large European organisation. I've seen what survives contact with real customers, and what quietly gets switched off.

Now I build the same thing for Belgian companies, in French, Dutch and English.

See my background on LinkedIn

Not ready to talk yet?

Then just tell me what eats your week.

Three things, in rough order. No need to think in terms of AI, that part is my job. I'll tell you honestly whether it's worth automating, and whether it needs an agent or just a workflow. No deck, no pitch.

Connecting the dots

Start with one workflow. See how fast it compounds.

You don't need an AI strategy. You need one process running end to end, and a platform that lets you add the next ten without starting over.