A 14-store import group calls with a specific request: automate Google, leave Meta alone. Their social specialist has run that channel for six years, knows every F&I manager's name, and produces creative the dealer principal actually likes. The group doesn't want a platform that automates everything or nothing. They want one channel handed over and one channel left exactly where it is.
That request is becoming the default shape of the conversation, not the exception. And most of the platforms fielding it can't actually do what they're describing.
Is Dealership Marketing Really an Either/Or Between In-House and Agency?
The framing that dominates search and sales conversations right now is binary: dealership-run marketing versus agency-run marketing, pick a side. It's a clean pitch for both camps — the agency selling full-service management, the software vendor selling full automation — but it doesn't match what a multi-rooftop group's marketing org actually looks like on the ground.
A 20-store group doesn't have one marketing posture. It has a used-car director who trusts a specific Meta rep because that rep found $40,000 in wasted spend last year. It has a Microsoft account nobody has opened in eight months because nobody on staff understands Audience Network. It has a Google account that's been running the same Search campaigns since 2019 because the person who set it up left and no one has touched it since.
That's not a company deciding between two philosophies. That's a company with four or five genuinely different channel situations, each deserving a different answer. Per-channel is not a compromise position between full automation and full agency control — it's the accurate description of where most groups already are.
Why Do Dealer Groups Want a Hybrid Automation Model?
Three reasons show up constantly when this comes up in practice.
First: sunk trust. A channel with a specialist who's already earning results is not a channel a GM wants to hand to software, however good the software is. That trust took years to build and the dealer isn't interested in re-litigating it to save a management fee.
Second: contractual reality. Plenty of groups are mid-contract with an agency on one or two channels and have no interest in breaking that relationship early, even while they're ready to automate everything else.
Third — and this is the one that gets underweighted — commodity-channel fatigue. Search bid management, PMax feed hygiene, Demand Gen creative refresh: these are execution tasks with a right answer, not a taste question. Groups increasingly see no reason to pay a human $3,000 a month to do something that's fundamentally mechanical, while they're still willing to pay for judgment on the one channel where judgment is actually differentiating.
Put those three together and you get a genuinely rational ask: automate the commodity execution, keep the relationship-driven channel where it is. Not agency-vs-platform. Channel-by-channel.
Why Can't Most Ad Platforms Actually Honor a Per-Channel Opt-Out?
Here's where the request runs into a wall. Most platforms that market "per-channel control" mean something much smaller than the dealer thinks they mean.
In practice, "per-channel" in most ad-tech stacks means a toggle that stops new campaign creation on a channel. It does not mean the platform's optimization engine, its budget model, or its reporting stop reasoning about that channel. The budget balancer still sees the channel's historical spend when computing pacing targets. The performance dashboard still nets the agency-run Meta account against the platform-run Google account when it reports blended CPL. The nightly rebalance job still treats the whole account as one pool of dollars, because that's how the system was built — one advertiser, one optimization objective, channels as sub-line-items inside it.
That's not malicious. It's architectural. Most ad-tech platforms were built assuming they own the full account. Retrofitting a genuine boundary — where an entire channel is invisible to the system's own reasoning, not just closed to new campaign launches — means rewriting the run router, the budget matrix, and the spend ceiling to treat that channel as outside the system's universe entirely. Most platforms treat "per-channel" as a UI feature. Structurally, it's a much deeper claim than a toggle can deliver.
The dealer asking for "keep Meta in-house" isn't asking for a UI toggle. They're asking whether the system can genuinely not know that channel exists — can't launch on it, can't analyze it, can't count it against a shared ceiling, can't retroactively reason about performance on campaigns it never built. Most platforms answer yes and mean "we'll stop touching it going forward." That's a materially smaller promise.
What Does "Genuinely Untouched" Actually Require?
A real per-channel boundary has to hold at three separate points, not one.
It has to hold at the point of launch — the system refuses to create anything on the opted-out channel, obviously. But that's the easy part and the part every vendor already claims.
It has to hold at the point of analysis and rebalancing — the system's daily reasoning about where budget should move can't quietly use the opted-out channel's performance data as an input, and can't shift dollars into or out of it as a side effect of optimizing everything else.
And it has to hold at the spend ceiling itself — the dealer's daily or monthly cap has to exclude that channel's spend entirely, including spend on campaigns the platform's own AI never built. If an agency is running $8,000 a month on Meta and the platform's spend ceiling logic nets that against the dealer's approved budget, the dealer hasn't gotten a real opt-out. They've gotten a shared budget with an asterisk.
AUTONOMi shipped a per-channel dealer opt-out in July 2026 that a dealer can flip for any channel — Google, Meta, TikTok, Microsoft, CTV/Streaming, or Fixed Ops — handing that channel back to themselves or their existing agency relationship with one switch.✓ Jul 22 The opt-out is enforced at the run router, the governance layer, the budget matrix, and the spend ceiling simultaneously, so an opted-out channel is excluded from every layer that could otherwise touch it — not just blocked from new campaign creation.✓ Jul 22
What Happens to AEGIS's Own Campaigns When a Channel Is Opted Out?
The mechanical detail matters here, because it's the part most vendors gloss over. When a dealer opts a channel out, does the automation's own prior work on that channel just sit there, orphaned, still spending against a budget nobody's tracking?
On opt-out, AEGIS pauses its own campaigns on that channel — it doesn't leave them running unmonitored. Dealer-owned or agency-run campaigns on that channel are never touched by AEGIS, and their spend is excluded from AEGIS's own budget ceiling entirely.✓ Jul 22 The channel's freed-up allocation redistributes to the channels AEGIS still manages, rather than sitting idle or silently padding a blended average.
That last point is the one that actually answers the dealer's original question. "Keep Meta in-house" doesn't just mean "don't launch anything new there." It means the dealer's existing Meta specialist, or their agency, keeps full and undisturbed control — no shadow optimization, no retroactive performance claims layered on top of work they didn't do, no budget math that quietly assumes visibility into an account the platform was told to leave alone.
Does the Commodity/Judgment Split Actually Hold Up Channel-by-Channel?
It's worth being honest about where this cuts against the dealer's instinct, too. Not every channel splits cleanly into "commodity execution" and "relationship-driven judgment," and the line moves depending on the store.
Search bid management on Google looks purely mechanical until an OEM launches a regional incentive mid-month and someone needs to catch it fast. PMax feed hygiene looks purely mechanical until a trim mismatch quietly tanks Vehicle Listing Ads performance for three weeks because nobody was watching the feed. The commodity/judgment line isn't fixed — it moves with how much attention a channel is actually getting today, in-house or agency-run.
That's the actual argument for genuine channel-level opt-out over an all-or-nothing decision: the dealer should get to reassess per channel, on their own timeline, without the switch being a one-way door. A specialist leaves, a channel that was "judgment" becomes "commodity" overnight — the dealer should be able to hand it to automation the same week, not renegotiate a whole-account contract to do it.
The AUTONOMi Approach to Hybrid Channel Control
AEGIS runs a single daily allocation decision across every paid sub-channel it manages, made as one reasoning pass rather than piecemeal per-channel calls — which is exactly the architecture that makes a genuine opt-out possible instead of merely cosmetic.✓ Jul 22 Because the allocation engine already reasons about the whole managed portfolio as one system, removing a channel from that system means the channel is actually gone from the reasoning — not silently retained as a background input.
The per-channel opt-out covers Google, Meta, TikTok, Microsoft, CTV/Streaming, and Fixed Ops individually — a dealer can hand back any one of them while AEGIS continues to run the rest, with the opted-out channel's budget redistributing to the channels still under management.✓ Jul 22 This is distinct from the allocation locks AEGIS also supports, where a dealer holds a specific budget split fixed but AEGIS still executes on that channel — opt-out is a harder boundary: AEGIS doesn't execute there at all.
Every allocation shift, lock, and opt-out is hash-chained into the dealer's own audit trail, so the dealer can see exactly when a channel was handed back and confirm nothing was launched against it afterward.✓ Jul 22 That auditability is what turns "we promise not to touch it" into something a CFO can actually verify rather than take on faith.
The result is a dealer group that can run a genuinely mixed model — an in-house Meta specialist, an agency handling Fixed Ops, and AEGIS running Google, TikTok, and Microsoft — without any of those three parties stepping on the others' budget, reporting, or campaign history.
What Should a Dealer Group Ask a Vendor Before Trusting a Per-Channel Claim
The next time a platform says it supports per-channel control, the question worth asking isn't "can I turn a channel off." It's narrower and more useful: when I turn it off, does your system stop counting that channel's spend against my budget ceiling, stop using its performance data in your optimization, and stop reasoning about it at all — including retroactively, on campaigns your system never built? Most vendors will answer the first question confidently and go quiet on the other two, because the other two require an architecture most of them don't have.
That's the actual dividing line in this market now — not dealership versus agency, but which platforms built channel boundaries deep enough to survive a dealer changing their mind. Groups running a hybrid model today can model what a genuine per-channel split would do to their current spend before deciding which channel, if any, they're ready to hand over next.
