It recommends. Your platform decides.
AURA Route scores every inbound contact against your own historical first-contact-resolution data and tells your routing platform which destination is most likely to resolve it. Your platform is free to take the advice or ignore it.
Sits alongside Genesys, Avaya, NICE CXone, Talkdesk, Five9 or Amazon Connect. No migration.
Why it doesn’t route anything itself
AURA Route has no execution rights on your contact centre. It receives a contact event, produces a recommended destination with a confidence score and a stated reason, and returns it. Your routing platform then does whatever your routing platform is configured to do.
You can wire that recommendation in as a hard route, as one input among several, as an override that only applies above a confidence threshold, or as nothing at all while you watch it for three months in shadow mode. That decision stays in your Architect flow, your routing rules, your change control.
This rules out a class of failure that autonomous routing cannot rule out: AURA cannot mis-route your contact centre at three in the morning, because AURA cannot route your contact centre at all.
The question your risk team will ask is what happens when the model is wrong.
The answer is that your existing routing rules happen — exactly as they do today. That is the entire safety story, and it is why there is no autonomy roadmap.
Shadow mode is the default starting point. AURA scores every contact and logs what it would have recommended, while your current routing runs untouched. After a period of your choosing you compare the two on your own FCR data, and decide whether it earns any influence at all.
Five stages, one recommendation
Every stage exists because it measurably improves the recommendation. Nothing here is AI for its own sake.
-
Ingestion
Voice, chat, SMS, email and social contacts enter one normalised pipeline. AURA doesn’t care what platform is underneath — it reads events, it doesn’t take over.
-
Context enrichment
CRM records, prior interaction history, account status and sentiment signals are attached to the contact before any scoring happens.
-
Intent classification
A model classifies the contact’s intent and returns a confidence score. Low-confidence cases escalate to a deeper model rather than guessing. Your prompts, visible and editable.
-
FCR-weighted destination scoring
Each available destination is scored against your historical resolution data for that intent type. Not longest-queued, not most-available — most likely to actually resolve it.
-
Recommendation and nightly learning
The recommendation is returned with its reasoning and logged. Overnight, the model retrains on the day’s outcomes — including every case where your platform ignored the advice, which is often the most informative data you have.
Baseline before influence
AURA establishes your FCR baseline before any recommendation is allowed to affect a single contact. Every subsequent claim is measured against that baseline, in your own numbers.
We publish no benchmark uplift figures, because we don’t have any we can stand behind yet. The design partner programme exists to generate them honestly.
Full decision audit
Every recommendation is logged with its intent classification, confidence score, the destinations considered, the reason the winner won — and whether your platform followed it. Your compliance team will ask whether the decision trail is complete. It is.
What it needs from you
- To read
- Contact events from your platform, and resolution outcomes so it can learn.
- To integrate
- A webhook or data action. In Genesys, typically an Architect data action calling a synchronous lookup before queue assignment.
- Not required
- Platform migration, professional services engagement, systems integrator, or any change to your telephony.
You’ve heard an intelligent routing pitch before
Sits alongside
No rip-and-replace, no migration project, no six-month implementation. If your IT lead can configure a webhook, they can deploy it — and unplug it just as fast.
Learns your environment
Retrains nightly on your outcomes, not a generic industry average. The model that matters is the one fitted to your queues, your intents and your agents.
No SI channel
No professional services arm, no integrator partners, no implementation revenue. Nothing in the business model rewards making this complicated.
Advisory by design
It cannot act alone. Not a limitation we’re working around — the reason it’s deployable inside a change-controlled contact centre at all.
Transparent scoring
Confidence scores, destination rankings and reasoning are all exposed. No black box, and no “the model just knows”.
Priced by decisions
By routing decision volume, not per seat. Growing your headcount doesn’t grow your AURA bill.
By decision volume, not by seat
Three tiers by monthly routing decision volume. Annual agreements available. We’ll quote against your actual volumes rather than publishing a number that fits nobody.
Up to 75,000 decisions/mo
- Full five-stage recommendation engine
- Up to 4 enrichment sources
- Basic rules editor
- Standard MIS dashboard
- Weekly model retraining
- Email support
Up to 300,000 decisions/mo
- Everything in Starter
- Nightly model retraining
- Model registry and weight controls
- Prompt manager
- Up to 10 enrichment sources, 3 tenants
- API access
- Business-hours support
Up to 1.2m decisions/mo
- Everything in Growth
- Unlimited enrichment sources and tenants
- Dedicated onboarding engineer
- 24/7 critical incident support
- Bring your own model or key
- Multi-region deployment
- Custom retention policies
Three to five contact centres. Twelve months.
AURA Route is pre-general-availability. We’re looking for a small number of contact centres to run it — starting in shadow mode — and to be honest with us about what it gets wrong.
What partners get
- Full platform access at low cost for twelve months
- Direct influence on the roadmap through monthly reviews
- A model fitted to your environment from day one
- Founding-customer discount if you convert at month thirteen
- Case study rights — with your approval on anything published
What we ask
- Access to anonymised routing and resolution outcome data
- A 45-minute review call each month
- Honest feedback, including when it isn’t working
We review every application personally and reply within five working days — including when the answer is no. With three to five slots we’d rather be direct about fit than string anyone along.