LinkedIn
Most agencies will not tell you this: much of the work behind your PPC retainer has already moved into Google’s machinery. Bids, match-type structures, placements, ad combinations and now parts of targeting came off the manager’s desk in stages. The invoice rarely shows the change.
I used to tell clients that management meant hours. I was wrong about what those hours would continue to buy. Read the dates in order and the pattern is hard to miss: the mechanical work keeps shrinking, while the decisions get more consequential. A retainer priced around the old workload no longer describes the job.
Before 2016: The hand-built account earns its hours
Picture a $20k-a-month account in 2014. A manager builds separate keyword versions for exact, phrase and modified broad match. They adjust bids for mobile, late-night searches and visitors who reached the pricing page. They pull search terms and add negatives one by one. Some of that takes judgment. A lot of it is data entry with opinions.
That distinction matters because the data entry was part of what the retainer bought. Bid changes, match-type splits and ad copy tests produced visible work. More spend could mean more account structure to maintain, so charging by the hour or taking a cut of spend had an intuitive logic.
Keep that workload in view. The dates that follow do not show Google making PPC judgment unnecessary. They show Google taking over tasks that once made the judgment look labor-intensive. If someone still prices management as though every bid and keyword variation needs a human hand, this is the account they are asking you to pay for.
2016 to 2018: Smart Bidding takes the bid lever
Target CPA and Target ROAS did not merely suggest bids. They set them auction by auction, using signals such as device, location, time and past behavior faster than a person could work through a bid table. That changed the manager’s task. Instead of changing hundreds of bids, they had to supply clean conversion data and choose a target the system could reasonably pursue.
Bad inputs did not become good decisions just because an algorithm made them quickly. If the tracking counted the wrong action, or the target pulled the campaign away from valuable customers, automation could repeat that mistake at scale. The human responsibility moved upstream.
What shrank was the bid-changing routine, not accountability for the bid strategy. The weekly report listing manual adjustments became a weaker explanation for a retainer. The harder questions were what counted as a conversion and what it was worth. For the mechanics of the options that remain, this guide to Manual CPC, Smart Bidding, tCPA and tROAS lays them out. If your manager bills for bid tweaks, ask which bids they changed by hand last week—and why.
February to July 2021: The old match-type build loses its purpose
In February 2021, Google folded broad match modifier into phrase match. By late July, advertisers could no longer create new +keyword variants. Exact match had already expanded to include close variants years earlier. This was not the day exact match stopped meaning exact; it was another reason to stop treating each match type as a rigid fence.
The old three-way build had been useful partly because it gave managers separate structures to tune. Once those boundaries changed, maintaining every variant and cleaning up duplicates became a less convincing use of paid hours. Google pointed advertisers toward consolidation or broad match with Smart Bidding, shifting more of the query-matching work into the system.
The remaining job was to decide where matching could roam. Negatives still mattered. So did the conversion signal used to teach broad match which searches were valuable. A huge keyword list could divide attention and signal without giving you the control its size implied. Before approving another expansion of the account, ask which theme needs its own treatment and which is just a duplicate wearing a different match type.
November 2021: Performance Max takes placement choice out of the build
Performance Max launched in November 2021 to replace Smart Shopping and Local campaigns. One-click upgrades began in April 2022, and forced auto-upgrades were completed by September 2022. A single campaign could run across YouTube, Search, Display, Discover, Gmail and Maps. Instead of building a separate placement plan for each surface, advertisers supplied assets, feeds and a goal; Google decided where impressions served.
That removed a familiar kind of control. It also made poor inputs harder to hide. A feed title that misses search intent, creative that fails when cropped or a conversion value that treats every sale alike can steer automated delivery toward the wrong result. The manager cannot fix those problems by polishing a channel-by-channel bid sheet.
Inputs became the practical control surface. That does not mean every concern about a placement can be solved with a better image or a feed edit. It means a placement-pruning report, on its own, no longer demonstrates that someone is managing the campaign well. Ask instead what the campaign was told to value, which assets it had to work with and what the resulting traffic was worth.
June 30, 2022: Fixed search ads give way to assembled combinations
Expanded text ads lost creation and editing on June 30, 2022. Responsive search ads became the search ad type advertisers could create. You could still write headlines and descriptions, but you no longer wrote one fixed Headline 1, Headline 2 and Description 1 unit. You supplied up to 15 headlines and 4 descriptions for Google to assemble.
That is a meaningful difference for anyone billing for line-by-line ad tests. The neat A/B story—version B won because of one changed word—no longer describes most of the work. Google’s figures on that page claimed 7% more conversions at a similar CPA for advertisers switching to responsive search ads, and 20% more when paired with broad match and Smart Bidding. Those claims do not write your offer for you.
The manager still owns what every usable combination promises. Google can rearrange headlines; it cannot decide whether free shipping, same-day booking or a particular price is a real reason to buy. Nor can it repair a landing page that contradicts the ad. Stop judging the work by the number of headlines uploaded or CTR alone. Check whether the possible messages remain true and specific after assembly.
October 2023: Demand Gen changes the creative assignment
Demand Gen became available in October 2023, when advertisers also lost the ability to create Discovery campaigns. It later absorbed YouTube Video Action campaigns. From a distance, the change could look like a rename. Up close, the creative assignment widened: Discovery allowed image, carousel and feed creative, while Demand Gen added YouTube video and Shorts, lookalikes, and Max Clicks bidding.
The manager was not relieved of creative work. They were relieved of the comfortable idea that a static feed campaign could be set up and checked once a month. Reusing a display banner does not make it native video. Moving an old asset into a new campaign is an upload, not a prospecting strategy.
The decision is what creative deserves distribution, and how you will judge it. If your manager launched Demand Gen by copying Discovery assets across, ask what changed to suit the new surfaces. Look at new-customer rate and how often the creative is refreshed, rather than treating blended CPA against Search as the whole verdict. The campaign name changed; the need for a deliberate offer did not.
March 2025: Enhanced CPC loses its middle ground
Enhanced CPC stopped being available for Search and Display effective the week of March 31, 2025. Campaigns that were not migrated now run as plain Manual CPC. ECPC had been a compromise: a person set the bid, then Google adjusted it up or down for an auction. Removing it made the choice starker.
A manager could use Smart Bidding and take responsibility for the conversion data and target, or use Manual CPC and own the bid work without that auction-time adjustment. Neither route is a substitute for understanding the account. Manual control can still serve an edge case, including one without enough conversion volume to support the automated approach. It is not automatically evidence of more attentive management.
What disappeared was the half-manual default, not the need to choose deliberately. If an account is still on Manual CPC, ask why control is worth the ongoing work. If it has moved to Smart Bidding, ask what the system is optimizing toward. A retainer should pay for that reasoning, not for the comforting appearance of a person touching a bid column.
May 2025 to April 2026: AI Max pushes beyond the keyword list
AI Max for Search entered open beta in May 2025 and reached general availability on April 15, 2026. It is not a new campaign type. It is a suite within Search that can expand search-term matching, customize text with generated headlines and expand the final URL. Google can take a larger role in deciding which query, message and page meet each other.
That changes the meaning of a keyword list. It remains an input, but it is less of a complete instruction sheet for where traffic goes. A manager who treats AI Max as a toggle has skipped the difficult part: deciding how far expansion can go without misrepresenting the offer or sending a promising search to the wrong page.
Guardrails become the work. Brand exclusions, URL expansion limits, useful creative assets and a site structure that gives expansion somewhere sensible to land all matter. So does checking where the traffic actually went. In the first 14 days, inspect the queries and destination pages before deciding what to allow next. The question is not whether someone clicked the switch. It is whether they noticed what happened after they did.
June to September 2026: Automation reaches budget tolerance and legacy campaigns
On June 15, 2026, Google announced three bidding and budgeting changes: a Promotion Mode beta for scheduled ROAS tolerance and a budget boost; Smart Bidding Exploration expanding from Search to feedless Performance Max globally, alongside a Shopping beta; and an August 17 backend change affecting budget-limited Target CPA and Target ROAS campaigns that beat their targets. The published example makes the last change plain: a campaign targeting $10 and achieving $5 could have bids pushed back toward the $10 target.
That is a different kind of automation from saving someone a bid-table chore. It changes how the system uses room between a target and actual performance. Beating a target by half is not, by itself, proof that the campaign will keep operating at that level while budget is constrained. Someone has to decide whether the extra volume is worth the movement toward target.
Google also split the AI Max cleanup. Automatically created assets and campaign-level broad match were set to auto-upgrade in September 2026, while Dynamic Search Ads automigration was delayed on June 11 from September 2026 to February 2027. DSA creation was restored until January 2027, and its migration carries the most aggressive default, with all three AI Max features switched on.
The dates matter because the defaults are not identical. Check budget-limited campaigns after August 17 before celebrating a CPA far below target. For legacy campaigns, check what is scheduled to move and which controls will be on when it does. That review is real work. Waiting for the migration email and forwarding it to a client is not.
September 2026: The old retainer meets the new job
Put the changes beside the work they displaced. Auction-time bids moved into Smart Bidding. Match-type maintenance lost much of its old purpose. Performance Max reduced placement choice. Responsive search ads assembled copy combinations. Demand Gen changed the creative brief. ECPC’s compromise disappeared. AI Max widened matching, text and URL expansion; newer changes reached budgets and legacy campaigns.
The mechanical work did not vanish all at once. It stopped being a sound basis for pricing the whole job. Managers still make consequential decisions: what counts as a conversion, what it is worth, which traffic to exclude, which offer to put in front of a customer and which pages should receive the visit. Those decisions are harder to display as a long activity log. They also matter more than a long activity log.
The billing model has been slower to change. A 2026 pricing guide puts typical management at $1,500 to $5,000 a month for $5k to $25k of spend, with percentage-of-spend fees commonly at 10% to 20%. Its example is simple: 15% of $20k is a $3,000 fee. The guide also notes where that model weakens: spend can grow without the workload growing alongside it. The timeline shows why that question is worth asking before signing another contract.
I would buy management against a written outcome, not an inherited task list. Set a 90-day block, a baseline for cost per acquisition or return on ad spend, and, for B2B, a baseline for qualified pipeline. Agree on the target and guardrails. Then inspect the changes made and the result they produced. If the number does not move, hours in a report do not make the fee worthwhile.
That is why I run accounts through groas now: machines handle daily execution and log changes, while a strategist owns targets and guardrails with me. It is a clearer split between what automation does and what a person must answer for. If your next contract cannot name the number, the baseline and the date, do not sign it.