Rethinking Routing in Modern Support

# SuccessLab Roundtable

From Static Rules to Continuous Decisions

February 10, 2026

At the SuccessLab roundtable on AI-enabled triage and routing in San Mateo, routing quickly became its own conversation. It was discussed as a decision discipline, not a configuration task or queue-management problem.

Most organizations assume routing is solved.

A case arrives. Rules apply. The case lands somewhere.

The leaders in the room challenged that view.

What emerged was straightforward and uncomfortable. Routing is no longer a one-time operational step. In modern support environments, it must operate as a continuous decision process, adjusting as understanding improves.

AI is accelerating routing, but more importantly, it is revealing how fragile many existing routing models are. One VP put it bluntly: "We optimized routing for our dashboards, not for our engineers. Guess which one broke first?"

Reframing Routing

Early in the discussion, the group aligned on a simple distinction.

Routing is not about where a case starts. It is about where it should go next.

Most routing systems assume the initial understanding of a case is correct. In practice, intake is incomplete. Context evolves. Impact becomes clearer. Risk often surfaces later.

As one leader observed, “teams route based on their early interpretation of a case, and that interpretation frequently changes.”

Routing systems that cannot adapt to this reality rarely fail at intake. They fail later due to friction, rework, and escalation.

Why Static Routing Breaks

Across organizations, the failure patterns were familiar.

Routing logic is usually driven by:

Initial categorization

Product or component tags

Customer tier

Basic skill labels

These inputs are treated as facts. In reality, they are starting assumptions, and often wrong ones.

Here's an example: "A customer reports "slow performance." Intake tags it as infrastructure. Routes to platform team. Platform team investigates and discovers it's a query optimization issue. Routes to database team. Database team finds a bad join pattern in application code. Routes to dev team. Three reassignments. Two days elapsed. Customer still waiting."

Understanding improves as:

Additional context is gathered

Telemetry and change events surface

Engineers engage

Customers clarify business impact

When routing does not adjust to this learning, cases bounce, stall, or escalate. Leaders described this as “thrash rather than volume”.

The real cost is not reassignment itself. It is the confusion created when direction keeps changing without explanation, and the erosion of trust when your best people start gaming the system to avoid bad routes.

What Routing Is Actually Optimizing For

One of the most debated questions was deceptively simple. What does good routing optimize for?

There is no single answer. And that's where most organizations fail, they pretend there is. Routing decisions may prioritize:

Speed to resolution

Depth of technical expertise

Familiarity with a customer environment

Load balancing across the team

Learning opportunities for developing engineers

Problems arise when these priorities are implicit rather than intentional.

If a team cannot explain why a case was routed a certain way, it does not control its routing model. The model is controlling them.

One VP of support operations described it this way: "We ran a post-mortem on a P1 that bounced four times. When I asked why it kept moving, nobody could tell me. The system just... did it. That's when I knew we'd lost control."

AI forces this clarity by making tradeoffs visible.

Dynamic Routing and Guardrails

There was broad agreement that AI enables dynamic routing. There was equal agreement that dynamic routing requires boundaries

Allowing cases to move as understanding improves is healthy. Allowing them to move endlessly is not.

Leaders aligned on practical guardrails:

Routing changes must be explainable

Confidence must be visible

Human override must be immediate

Repeated re-routing should trigger review

But here's where it got tense: How many moves is too many?

"Two moves? Three? Five? We didn't land on a number because context matters. A complex infrastructure incident might legitimately need four routing decisions as the failure domain becomes clear. A billing question that moves three times is a broken intake."

Understanding evolves. Routing should evolve with it. Chaos is optional.

Preventing Concentration Risk

A critical discussion focused on unintended consequences.

AI routing systems tend to favor top performers. They resolve cases faster, escalate less, and appear efficient in dashboards.

Over time, this creates single points of failure.

This is not just a workforce issue. It is a resilience issue.

One participant shared: "Our AI kept routing the hardest problems to our principal engineer. Great for MTTR. Disaster when she went on vacation. We had sixteen P1s stall because nobody else had context on those systems."

Leaders emphasized the need to:

Monitor concentration of complex work

Cap exposure to high-risk cases

Distribute learning opportunities deliberately

Design routing to build capability across the team

Good routing strengthens the system instead of exhausting a few individuals. Most routing algorithms optimize for efficiency. The best ones optimize for resilience.

Trust and Explainability

Routing only works when people trust it.

Trust erodes when cases arrive without context, when decisions feel opaque, and when mistakes cannot be corrected in the moment.

Modern routing systems must be able to answer:

Why this case came here

What signals were used

How confident the system is

What to do if the route is wrong

If agents cannot flag a bad route easily and see that feedback influence future decisions, they will work around the system. And they are creative about it.

Shadow Slack channels. Private case notes. Informal handoffs that bypass the queue entirely.

Routing rarely fails loudly. It fails quietly when belief disappears.

Metrics That Matter More Than Speed

Traditional routing success is often measured by speed and SLA compliance. The leaders challenged this framing.

More meaningful indicators surfaced:

Reassignment frequency

Escalation volatility

Time to correct an initial route

Confidence at the decision point

Distribution of complex work across the team

If routing improves dashboards but increases frustration on the floor, it is not working.

Here's a metric worth watching: reassignment within the first hour. If you're reassigning 40% of P1 cases in the first sixty minutes, you don't have a routing problem. You have a triage problem. Fix upstream.

Where Leaders Still Differed

There were healthy differences of opinion, and some sharp ones.

Some leaders favored more aggressive AI-driven routing early in the lifecycle. Others preferred tighter controls in regulated or high-risk environments.

The sharpest disagreement was about autonomous re-routing. Should AI be allowed to move a case that's already been assigned?

The aggressive position: Yes, if confidence improves. Why leave a case with the wrong team just because we made an initial decision?

The conservative position: No, unless critical. Every reassignment burns trust. Make the first route count.

There was also debate about how visible routing logic should be to customers. Should they see that their case moved teams three times? Or is that internal operational noise?

We didn't resolve it. Context matters too much. A fintech company dealing with regulatory scrutiny has different transparency obligations than a B2B SaaS platform.

These differences reflected context rather than disagreement on core principles.

What the Group Agreed On

By the end of the session, several conclusions held:

Routing is a continuous decision process

Static assumptions do not scale with complexity

Tradeoffs must be intentional

Dynamic routing requires guardrails and transparency

Concentration risk is a design responsibility

Trust depends on explainability and correction

Routing is where intent meets reality. When it breaks, everything downstream absorbs the cost.

Final Reflection

Routing used to be operational plumbing. AI is turning it into a strategic lever and exposing every assumption we buried in our CRM workflows.

What stood out was not the technology. It was the focus on adaptation rather than speed. Leaders were less concerned with how fast cases move through queues and more focused on how deliberately work is redirected as understanding improves.

Routing is no longer about distribution. It is about leadership decisions, encoded in systems.SuccessLab Roundtable, San Mateo, CA January 29, 2026

SuccessLab Roundtable, San Mateo, CA, January 29, 2026

Comments (0)

Popular