Key takeaways
- Most new AM ramp time is archaeology, reconstructing decisions that were never written down, not actual training.
- A useful handover captures reasoning, not just structure: why a campaign is paused matters as much as the fact that it is.
- Versioning decisions with dates and sources lets a new AM answer why questions without booking time with the outgoing AM.
- Context classified by recency and business impact is more useful for onboarding than a complete but unranked archive.
- Manual capture helps but stays incomplete. A system that captures context automatically keeps the record current as the account evolves.
Most new account manager ramp time is not training time. It is archaeology time, spent reconstructing decisions that were never written down.
The average new AM at a paid media agency spends two to three weeks asking the same category of question: why is the account structured this way, what did we test last quarter, why is this campaign paused, who approved this budget split. The answers exist somewhere, split across the outgoing AM's memory, a Slack thread from six months ago, and a Google Doc nobody has opened since Q2. Getting to them takes calendar time with people who are already overbooked.
This guide is about fixing that, practically, without a six-month documentation project.
What a new AM actually needs to know per account
Before you can capture context well, you need to be specific about what context matters. For a paid media account, the useful onboarding knowledge falls into roughly four categories.
Account structure rationale
Why campaigns are organized the way they are. Which segments get their own campaigns versus ad groups, how budgets are split between brand and non-brand, why certain match types are used or avoided. This is almost never written down because it felt obvious at the time.
Test history
What has been tried and what happened. Not just the wins. The things that were killed matter as much or more, because a new AM who does not know about them will propose running them again. A rough estimate based on mid-size agency patterns: teams re-propose previously killed tests somewhere between once and three times per quarter per account. Each re-proposal costs a meeting and sometimes a wasted flight test.
Active decisions and constraints
What changed recently and why. Budget shifts, paused campaigns, creative pulled by Legal or brand, KPI changes from the last client QBR. These are the decisions that will make a new AM's first recommendations look out of touch if they do not know about them.
Client context
How the client prefers to communicate, what they are sensitive about, which metrics they actually watch versus which ones are in the report because someone asked for them once, and any history that affects the relationship.
Why the standard handover document fails
Most agencies handle AM transitions with a handover doc. The format varies but the problem is consistent: the doc captures a snapshot, not a living record, and it is already partially stale the day it is written.
The outgoing AM writes what they remember to write, which skews toward structure and away from reasoning. Why a campaign is paused does not make it into the doc as often as the fact that it is paused. The Slack thread where the client asked you to pull spend from that ad set, and the three follow-up messages that explain the business reason, stays in Slack.
The new AM reads the doc, has questions the doc cannot answer, and books time with the outgoing AM or the account lead. If the outgoing AM has already left, that path is closed. The institutional knowledge left with them.
This is not a discipline problem. It is a system design problem. Asking people to document reasoning in real time, on top of their actual job, produces incomplete documentation. The incentive to write it down is low when you are the one who already knows it.
A practical framework for capturing context once
The fix is to make capture a byproduct of work that is already happening, not a separate task.
Connect the sources where decisions actually live
For most paid media teams, decisions happen in four places: ad platform accounts, Slack, meeting recordings and notes, and documents like briefs or strategy decks in Notion or Google Drive. A new AM needs access to all four, but raw access is not the same as useful access. Reading six months of Slack history is not a realistic onboarding activity.
The goal is to make the relevant signal from those sources searchable and attributed. When a new AM asks why the branded campaign budget was cut in March, the answer should surface the Slack message where the client made the call, the meeting note where it was confirmed, and the date it took effect in the account. All three, connected, in one place.
Version decisions as they happen
Every significant account decision should be versioned with a date, a reason, and a source. Not a formal change log that requires a separate workflow, but a lightweight record that is created automatically when the decision happens. Budget changes in the ad platform, strategy changes in Slack, creative decisions in email or meetings.
Versioning matters for onboarding because it answers the when and why questions that a static doc cannot. It also matters for continuity: if a client disputes a decision six months later, you have a record of what the agent knew and when.
Prioritize by recency and business impact
Not all context is equally useful for a new AM. A test result from three years ago is less relevant than a budget decision from last Monday. A throwaway Slack message from a junior team member is less relevant than a strategic direction set by the client's CMO.
A useful memory layer classifies context by recency, source authority, and business impact before surfacing it. This is what separates a searchable archive from a useful onboarding tool. The archive has everything. The onboarding tool shows the new AM the three things that matter most this week, not fourteen things of equal apparent weight.
How this changes the first two weeks
With a context layer that is current and prioritized, the onboarding sequence changes in a specific way.
Day one through three can be spent on the ad accounts directly, reading structure and performance data, with the reasoning available on demand. Instead of booking time with the outgoing AM to ask why a campaign is structured a certain way, the new AM queries the context layer and gets the answer with the source attached.
Days four through seven can shift to live account work, with the context layer catching cases where a proposed action conflicts with a recent decision. If the new AM drafts a recommendation to scale a campaign the team decided to sunset on Monday, the system flags it before it goes to the client.
Week two can be client-facing, because the new AM has enough account history to have informed conversations rather than spending client time asking background questions that should have been answered internally.
The three-week archaeology phase compresses to a few days of directed review, because the answers to the why questions are findable rather than buried.
What to do before you have a memory layer
If you are running a manual process today, the highest-leverage thing you can do is capture the reasoning behind decisions at the moment they are made, not after the fact.
When a campaign is paused, write one sentence in Slack tagging the account channel: what was paused, why, and who approved it. When a test is killed, add a line to a shared doc with the result and the reason it was ruled out. When a client changes a KPI or constraint, note it in the meeting summary with the date.
This is not a complete solution. It still requires discipline and it still produces a partial record. But it is dramatically better than relying on the outgoing AM's memory, and it gives a new AM something to search when they have questions.
The limitation is that manual capture stays incomplete over time. The team captures decisions when they remember to, which means the gaps accumulate. Peak season learnings from last year do not make it into this year's onboarding. The account lead who resigned took the reasoning with them even if they left a handover doc.
Where Almond fits
Almond is built as a Business Context Engine for performance marketing teams. It connects to ad accounts (Google Ads and Meta), Slack, Notion, Sheets, email, meeting notes and CRM in roughly four minutes, and captures decisions, tests and results as they happen without requiring anyone to write a brief or maintain a separate doc.
For agency onboarding specifically, the relevant capability is that context is versioned and retained over time, classified by relevance, recency and business impact, and scoped per client so one account's history cannot surface in another's recommendations. When a new AM asks why a campaign is structured a certain way, the answer comes with the Slack thread and the change log attached.
Every piece of context is auditable: you can see what the agent knew and when it knew it, which matters both for onboarding and for client conversations where decisions get questioned later.
Almond is in early access. Each member gets a detailed one-on-one Meta and Google ad account audit with the CEO, recorded. The waitlist is at almond.ai and requires a work email.
The core problem is not that agencies lack documentation discipline. It is that the system for capturing context was never designed to keep up with the pace at which context changes. A new AM should spend their first weeks learning how to apply your agency's judgment, not reconstructing it from scratch.
Frequently asked questions
- How long should it take to onboard a new account manager at a paid media agency?
- A well-documented account handover can get a new AM to independent client-facing work in five to seven business days. Most agencies take two to three weeks because account reasoning lives in people's heads rather than in a searchable system. The ramp time is mostly archaeology, not learning.
- What should be included in a paid media account handover document?
- Campaign structure rationale, test history with results and reasons tests were killed, active constraints like paused campaigns or legal-flagged creative, recent client decisions that changed KPIs or budgets, and client communication preferences. Static docs miss the reasoning. Versioned records with source attribution are more useful.
- How do you prevent institutional knowledge loss when an account manager leaves?
- Capture decisions at the moment they happen rather than relying on exit documentation. Version budget changes, test results and strategic decisions with dates and sources. When context lives in the system rather than in a person, a departure removes the relationship but not the account reasoning.
- What is an agentic memory layer for performance marketing?
- An agentic memory layer reads the sources where marketing decisions happen, including ad platforms, Slack, meetings and documents, and retains decisions, tests and results in a versioned, searchable form. It keeps context current automatically rather than requiring manual updates, and prioritizes by recency and business impact.
- How do you stop a new AM from proposing tests the team already ran and killed?
- By maintaining a versioned test log that records not just outcomes but reasons. A new AM who can query test history and see that a specific approach failed on creative rather than targeting can propose a refined variant rather than a repeat. Without that record, re-proposals are almost inevitable.
Almond 2.0 early access
Get a 1:1 Meta and Google audit with the CEO.
Every waitlist member gets a 1:1 audit of their ad accounts. Work email only.