Introducing Iris 2.0

Today we shipped Iris 2.0, the second major release of our intake agent. It went out to every firm already running Iris the same day, with nothing for anyone to install.

The release is centered on how the conversations feel. An intake form asks its questions in a fixed order because a form has to. A conversation doesn't. The follow-ups in 2.0 land the way a careful listener's do, and the mornings got better with them.

What changed

Warmer conversation model. Questions arrive in the order the story needs them, not the order a script does. Underneath, that means tracking what's already been said and skipping or reordering whatever the caller has already covered, instead of working down a fixed list top to bottom. Callers finish more of what they started to say, and the extra detail they volunteer is exactly the detail attorneys use.

Qualification accuracy. Fewer weak cases reaching attorney desks marked qualified, and fewer strong ones read as cold. The threshold has always been the firm's — Iris doesn't set it and 2.0 doesn't change it. What's new is that each score now carries the reasoning behind it: what pushed a case toward qualified, what held it back, so a reviewer can check the call instead of taking the label on faith. Before any of that reaches a desk, a second, independent read of the transcript checks that the capture actually matches what was said, catching a premature or thin capture before an attorney ever sees it.

Spanish, designed natively. The Spanish conversation is not translated from English. Its question order, its pacing, and its disclosure were each written for Spanish first — not carried over from the English version and adjusted afterward. A caller who starts in one language and drifts into the other partway through is met in whichever language they're using, without restarting or losing anything already said.

Practice-area reporting. The firm dashboard now breaks volume and qualification rate down by practice area, with a month-over-month view next to both — the three things the reporting partners at multi-practice firms had been asking for by name.

Why this direction

The first months of Iris running on live firm sites taught us one thing repeatedly: the details that decide a case live in the part of the story people tell when nobody interrupts them. A form interrupts by design. Every field is an interruption, every dropdown a decision the caller didn't come prepared to make.

2.0 is built around removing those interruptions. That shows up in question order, in how long Iris waits before the next one, and in the summary an attorney reads the next morning.

The mornings got better with them.

What stayed the same

One conversation, one summary, one night shift at a time. The pilot terms, the pricing model, and the firm's ownership of its records are exactly what they were. So is the disclosure Iris gives at the start of every conversation: it introduces itself as the firm's intake agent, not as an attorney, before the first question is asked.