Ad Tech’s Communication Chaos Lives in the Deal, Not the Chat

Medialive AI: Fixing Ad Tech’s Communication Breakdown

Author:

Ara Ohanian

Published:

October 28, 2025

Updated:

November 20, 2026

In October 2025 AdExchanger profiled Medialive, a startup led by former Magnite chief revenue officer Joe Prusz, whose shared workspace lets the buyer invite a campaign’s DSPs, SSPs and publishers into chat rooms while AI reads the brief and flags setup errors. Prusz likened it to Slack or WhatsApp for media buying. A year on, we found no customer, funding or product announcement from Medialive in the trade press.

The problem remains. In its own survey of 2,020 marketers, published in October 2026, DoubleVerify found they spent 9 hours 58 minutes a week on manual optimization, 14 minutes less than in 2025, while 97% planned to use AI in some part of media planning, buying and optimization in 2026.

The chaos lives in the deal, not the chat

Most of the back-and-forth happens between platforms, where, as Tanja Mimica of the AI startup Kovva told AdExchanger in May 2026, “buyers are still in spreadsheets for most of the day.”

Most private deals are agreed by phone, Slack and email, and terms such as budget and brand safety settings never appear in the deal ID itself. “The aspects of that deal are not captured anywhere within the actual supply chain,” Kyle Dozeman, then PubMatic’s chief revenue officer, said in November 2025. The Trade Desk’s Will Doherty put the share of deal IDs that reach meaningful scale at about one in ten. Magnite’s Nick Allen said two-thirds of deals configured with the desired supply “deliver no or very little revenue.”

Summarize the conversation, or turn it into fields

Tools take two routes. The first makes the talk smarter. Medialive summarizes chats into a to-do list. Kovva’s agents, reached through Slack or email, check geos, pixels and pacing after launch and explain why two platforms report different numbers. Assistants such as Agentforce inside Slack follow the same idea.

The second removes the conversation. IAB Tech Lab’s Deals API, final as version 1.0 on February 6, 2026, lets an SSP push a deal’s dates, countries, floor price, curator and fee type into the DSP. The Tech Lab’s AAMP 3.0 draft, opened for comment in September 2026, routes each agreed line item to the Deals API, OpenDirect or OpenRTB, and only 3 of 71 fields need a round trip to the other side. Our explainer covers how buyer and seller agents talk over AdCP.

Automate the checking first, keep the commitments

A sensible order, as advice:

  1. Read-only checks: brief against setup, naming, post-launch QA, deals live but not bidding. Horizon Media was building pacing and naming-check agents in December 2025. A false alarm costs a minute.
  2. Explanations: why two platforms disagree, why a deal under-delivers. The agent drafts, a person decides.
  3. Drafts to sellers and clients, sent only once a person approves.

Keep people on anything that commits money or changes terms. One ad tech executive let an agent read insertion orders and launch campaigns, then shut it down after two days, asking, “What if it accidentally added a zero?” Yahoo DSP’s troubleshooting agent, for one, recommends fixes or executes them with approval. If you buy traffic to pages that earn from programmatic ads, as in content arbitrage, you sit on both sides of these threads, so start with the discrepancy check.

How to tell whether it saved any time

Not from vendor figures. Mediaocean says pilots of the Innovid agents in its NIVO AI delivered workflow efficiency gains of up to 90% over manual campaign setup, yet DoubleVerify’s survey shows manual hours barely moved in a year of rising AI use. Before switching anything on, log for a month the time from problem spotted to fix, messages per issue, days from deal agreed to deal bidding, and errors caught after launch. Then repeat it, as performance marketing tests any new tool. If the hours stay flat, the tool moved the work instead of removing it.

The chaos was never mainly a chat problem. It was terms living in messages, not in the systems that spend the money. A summary makes the thread searchable. A field makes it unnecessary. Automate the reading, standardize the terms, and keep a person on what commits money, whatever the big platforms’ agents promise.

Frequently Asked Questions

What is Medialive?

Medialive is an ad tech startup that AdExchanger profiled in October 2025 for its shared workspace for programmatic campaigns. Buyers open a room per brief and invite DSPs, SSPs and publishers, and the platform pulls campaign data through APIs, flags setup errors and answers status questions in plain English. By November 2026 we found no public customer or funding news, so ask for references.

Why do programmatic deal IDs fail to deliver?

The terms agreed between buyer and seller often never reach the systems that run the deal. Floors, budgets and brand safety settings are agreed in email or on calls, then entered separately on each side, so one mismatch can stop a deal from bidding. IAB Tech Lab’s Deals API sends the terms from the seller’s system to the buyer’s, removing the retyping.

What is the IAB Tech Lab Deals API?

It is a standard, final as version 1.0 since February 2026, that lets a seller’s system, usually an SSP, send a deal’s terms to a buyer’s system, usually a DSP, and check its status. It covers dates, countries, floor prices, guarantees, pricing type, inventory and curator fee type. It sits outside the auction, and a later version is meant to add negotiation.

Which ad operations tasks should stay human?

As a rule of thumb, anything that commits money or changes agreed terms: signing insertion orders, approving deals, moving budget, agreeing make-goods and messaging clients or sellers. Let an agent check, explain and draft, and keep approval on for the rest until its record in your account justifies more. Log what it did and what you approved, so mistakes can be traced and reversed.