Build an agency growth plan that does not break delivery

An agency can sign another client and become a worse business that month. The new contract brings revenue, but it also brings briefing, production, approvals, meetings and another set of promises the team must keep.

Before increasing acquisition, work out where that delivery will fit and what the account must contribute after its costs. If the answer depends on founders rescuing every project at night, the growth plan needs a delivery plan first.

Harry Long's account of expensive agency mistakes starts with retention, then moves through ownership, acquisition costs, contracts and adapting the offer. His longer agency-growth walkthrough adds partnerships and content. The useful lessons concern the operating decisions behind growth; his revenue figures are his reported experience, not a benchmark or a promise.

Find out why clients stay before trying to replace them faster

A long-term client usually requires less repeated acquisition work than a procession of short-lived accounts. That does not make every account worth retaining at any cost.

Review the existing book of work. For each client, record the agreed result or deliverable, fee, direct production costs, founder involvement, approval delays, unresolved problems and likely continuation. Include management time and revision work that never appears in the original estimate.

A client who pays on time can still be unprofitable if every task takes twice the planned effort. Another may be commercially sound but at risk because nobody explains what happens next. Those require different interventions: scope or pricing in the first case, better account management in the second.

Talk to the client before assuming the cause. Ask whether the work is helping, where expectations differ and which coming decision matters to them. Solve problems you own. Make client dependencies explicit. If access, assets or approvals are missing, show how that affects the schedule rather than letting the delivery team absorb the delay invisibly.

Harry describes frantic acquisition partly driven by his suspicion that existing clients would not stay. That creates a bad operating habit: more sales pressure conceals the delivery problems producing the churn. Fix the repeated failure before buying more opportunities to repeat it.

Some churn remains outside your control. In an agency diary, he describes losing a long-standing account because the client's business circumstances changed. A good relationship cannot remove that risk. Check revenue concentration and maintain a realistic replacement pipeline even when clients are happy.

Give each function an owner and each handoff a definition

Three founders can all work hard while leaving the same decision unmade. If everybody does a little marketing, sales, finance and delivery, nobody may know who owns a missed follow-up or an unstaffed project.

Harry recommends assigning functional responsibility. The practical version is to name the person who decides, the person who does the work and what evidence the next person needs. A small team can hold several roles without making all roles collectively ownerless.

For example, marketing can own generating and documenting relevant enquiries. Sales can own qualification and the commercial agreement. Delivery can own the planned output, and client success can own whether the work remains useful and understood. Finance needs a clear payment and margin view across them.

Use measures that connect rather than measures that encourage each function to optimize its own score:

Raw leads, signed clients and retention are useful measures only with shared definitions. A sales team can close unsuitable clients. A delivery team can retain accounts by giving away unsustainable amounts of work. Look at qualified demand, realized margin, client outcome and workload together.

A handoff should not depend on someone remembering a detail from a sales call. Preserve the decision where the next owner works, and make it possible to trace a consequential promise back to its source.

Before the client kickoff, prepare the decisions already agreed and the questions still open so onboarding resolves gaps instead of repeating the sales conversation. Harry's onboarding diary favors meeting face-to-face where practical; choose it when the relationship benefits, and offer remote participation when travel, convenience or accessibility makes that more useful.

Plan further ahead than the weekly task list

Weekly sprints help a team choose the next work. They don't show whether three client launches are converging on the same specialist next month.

In one diary entry, Harry explains that his team had weekly and monthly plans but needed client roadmaps looking beyond three months. The purpose was practical: let clients and the internal team see what was coming and resource it properly.

Use a longer plan to identify intended outcomes, major deliverables, dependencies and decision points. Keep uncertain dates visibly provisional. Then turn the approved near-term work into assigned tasks with realistic deadlines.

At a weekly review, compare the work sold with the capacity available. Ask which tasks are blocked, which approvals are overdue and which commitments need to be renegotiated. Name what will move if a new urgent request is accepted. A priority list that contains everything offers the team no usable priority.

Harry's newsletter and sprint diary describes the founders reviewing each client's weekly deliverables and the specialists required. That is a better planning input than simply counting how many new clients arrived.

Reserve capacity for revision, account communication and the occasional genuine incident. If the plan only works when every first draft is accepted and nobody is absent, it is overstated before the week begins.

Price the service you can repeatedly deliver

An attractive project can still be a poor foundation for predictable revenue. Harry's account of finding LinkedIn as an acquisition channel includes a shift from broad services toward quiz funnels, followed by content work. One difficulty with the funnel work was its project-based nature compared with an ongoing retainer.

A retainer helps forecasting only when there is continuing work the client values. Turning a one-off build into a monthly charge without a useful ongoing responsibility is not a durable solution. Conversely, a well-priced project with clear boundaries can be healthier than a retainer full of undefined requests.

Define the delivery unit. For a content service, that might be research, recording support, an agreed number of edited assets and defined revision rounds. For a funnel project, it might be discovery, question design, implementation, testing and handover, with ongoing optimization priced separately if needed.

Estimate the labour and specialist cost, then review the estimate against actual delivery. Include tools used for that client, account management and expected rework. Decide what fee and workload leave room for the business to operate. Growth that increases revenue while reducing the return on every hour needs a change in scope, price, process or customer fit.

If you offer a sample to win an early client, keep it bounded. Harry's first-client advice uses an honest, useful piece of free work to demonstrate ability. His “zero budget” title still leads to a paid software example. Even a manual sample consumes time. Set its output, usage rights, approval route and end point before treating it as economical acquisition.

Make the commercial paperwork explain the relationship

The contract should state what is included, what is excluded, what the client supplies and how changes are handled. It should also establish fees, payment timing, renewal or end dates, cancellation, notice and any agreed remedies. Use a suitable reviewed template rather than asking an AI model to improvise a new agreement every time.

Harry recommends a three-month notice period to make churn more predictable and give an agency time to address concerns or replace revenue. That is his commercial preference. The appropriate term depends on the engagement, buyer expectations and applicable law. Agree fair terms before work begins; a clause inserted merely to trap an unhappy client can damage the relationship it was meant to protect.

Be especially precise about outcome guarantees. “One hundred leads” can mean completed forms, people meeting a qualification standard, accepted sales opportunities or customers. “Booked meetings” can include no-shows and people the client cannot serve. Those are materially different promises.

If the offer includes a measurable commitment, define the event, qualifying conditions, measurement window, responsibilities and remedy. Don't guarantee revenue or a number of qualified meetings when the result depends on factors you don't control. Ambiguity is not an effective closing technique.

Payment processes should also be visible. Upfront invoicing can be appropriate for reserved capacity or a defined monthly service. Explicitly authorized recurring billing can reduce late-payment administration. It should come with clear amounts, invoices, renewal terms and a usable cancellation route.

The source sometimes frames automatic billing as giving a client less time to think about the fee. Reject that reasoning. The useful objective is reliable, agreed payment for understood work. Hidden charges or obstructed cancellation create a different problem.

Choose AI tools by the job they actually remove

Harry's agency AI walkthrough describes a connected stack for research, outreach preparation, creative production, reporting and client knowledge. His most useful boundary is retaining human responsibility for critical client communication, sales relationships and final strategic decisions.

Begin with a repeated task whose input, output and reviewer you can identify. “Use AI throughout the agency” is too vague to test. “Prepare a cited account brief from approved client records before the Monday review” is a job somebody can inspect.

A connector gives an assistant access to particular tools and actions. MCP means Model Context Protocol; it is a way of connecting capabilities, not proof that every app supports the workflow you want. Inspect what is available, what access it needs and what the person reviewing the output can verify. The research and approval workflow covers those distinctions in detail.

Keep a usable client knowledge workspace

Create a separate, permissioned collection for each client. Include the approved brief, audience, offer, brand rules, actual writing examples, goals, team roles, decisions, promises and relevant meeting notes. Keep dates and owners with information that can change.

With the appropriate consent and access, meeting recordings or transcripts can reduce repeated questions and help people who were not on the call. Harry mentions Fathom and other recording options. The job is preserving agreed context, not indiscriminately storing everything the client says forever.

A copywriter should be able to find the approved tone and offer. A designer should be able to check the current visual rules. An account lead should be able to distinguish a suggestion from a commitment. Have the assistant cite the underlying source and flag conflicting notes instead of smoothing them into an invented answer.

Limit access to the people who need it. Do not mix client records or feed confidential material into a service without the permissions and protections the engagement requires. A fluent answer is not proof that the assistant retrieved the correct client's facts.

Prepare and check outreach instead of surrendering it

AI can help clean a list, check a stated ICP, summarize public company context, draft an opening and organize replies. Review the identity match, evidence, exclusions and message before any external contact. A new role or funding event may justify a question; it doesn't establish a need or budget.

The source describes campaign creation becoming much faster through connected tools. His reported eight-minute setup time does not establish how long your process will take. A low setup time is not a useful win if the campaign contacts the wrong people or invents personalization.

Keep permissions for reading, drafting and sending distinct. Respect refusal and maintain suppression records. A connected application cannot create the authorization the operator lacks, and its pacing cannot guarantee protection from platform restrictions.

Produce creative options from a clear brief

Use AI to propose copy and design directions from authorized examples, brand constraints and real performance context. Harry describes feeding successful creative combinations back into the next round of production, sometimes using Canva templates through a connector.

Make the variations test different ideas. Forty nearly identical designs are not forty useful experiments. Check the claims, readability, rights and actual fit before publication. A prior winner is a starting hypothesis for another campaign, not proof that its imitation will work for another audience.

Verify reporting before acting or sending it

An assistant can summarize exported ad data or an appropriately scoped account connection. It can help locate wasted spend, compare periods and explain which result changed. It can also use the wrong date range, mix conversion definitions or calculate a persuasive-looking error.

Keep the reporting period, spend, attribution window, event definition and source visible. Reconcile consequential totals with the underlying records. A small change from a tiny sample shouldn't produce an automatic budget decision.

Review a client report before sending it, especially when the result is poor or requires a decision. Scheduled reports need agreed recipients, frequency and access. Automating delivery does not remove the agency's responsibility for the explanation.

Use content and partnerships without adding another delivery problem

Useful content can help prospects assess the agency and help existing clients understand its thinking. Harry reports that more active content periods coincided with stronger business performance. That is an observation, not proof that increasing post volume causes higher profit.

Choose a sustainable production system and a clear audience. His diary describes separate newsletters for clients, community members and prospects. Those groups have different relationships with the business. Give each a reason to receive the material, use appropriate permissions and avoid disclosing another client's confidential results to manufacture proof.

Founder and company content can also play different roles. A founder can explain a difficult decision; the company can show a documented method or permitted example. Coordinate them before creating several channels that compete for the same scarce attention.

Partnerships can add capability and introductions. A genuine service partnership, earned certification or documented shared result may help a young agency establish credibility. It needs to reflect work actually done. A logo or affiliate link alone does not make the agency an endorsed expert.

Harry also describes referral income from tools his agency uses. If you consider those programs, check the actual eligibility, commission basis, duration, payment conditions and termination terms. His claims about lifetime percentages and monthly earnings do not establish current terms or likely income for your agency.

Recommend the tool because it suits the client's problem. Disclose the commercial relationship clearly. Don't turn the client's software choice into a way to collect commission from an unsuitable purchase. Referral work can create support and maintenance obligations, so include those in its economics too.

Change the offer from evidence and protect time for the work

A pivot deserves a reason beyond a new tool being fashionable. Review requests from suitable prospects, delivery results, repeated objections, project margins and your team's actual strengths. Test a narrower change before abandoning a service customers still depend on.

Harry's own account moves through cold email, broader marketing, assessment funnels and content. The reasons include deliverability work, generalist competition, project revenue and concentration in a narrow category. Those are questions to examine in your business, not instructions to follow his service sequence.

Keep existing obligations visible while testing the new offer. Decide which work ends, which capability must be added and what evidence will justify more investment. Otherwise a pivot simply adds another service to an already overloaded agency.

The diary material supplies a few smaller operating choices worth keeping. A remote base should allow workable overlap with the team rather than shifting inconvenient meeting hours onto colleagues. A reliable desk and work environment can reduce friction. Harry's Tirana coworking account also suggests choosing a shared workspace or local professional community when working alone becomes isolating; weigh the useful contact against cost, travel time and privacy for client work. Protected blocks for focused delivery can help when paired with a clear route for genuinely urgent issues. These are working arrangements to test, not a prescription for a particular destination or a claim that a gym routine guarantees productivity.

Put the next prospective client into the actual plan before committing: named owner, suitable offer, expected margin, available specialists, client responsibilities and a start date the team can support. If one of those is missing, resolve it or change the promise. A sale is easier to celebrate than an honest capacity decision, but the capacity decision is what makes the sale deliverable.