A freelance writer let AI write his quotes for 3 months: what it changed, what it didn't solve

In brief: After three months of entrusting a significant portion of his quotes to AI, this freelance writer saved time on the first draft of the document, but not on proofreading. AI for writing a quote speeds up formatting and structure, without replacing detailed knowledge of each client or tariff verification. This finding is based on a real-world test, conducted on several dozen quotes, using different tools and prompts. Here's what really changed in his daily life, and what continues to take up his time each week.

Why this test? How was it conducted?

The starting point is simple. This freelance writer, specialising in web content and copywriting, spent several hours each month drafting his quotes. Not the prices, which were already fixed. The part that took up time was the description of services, the terms, and the formatting adapted to each prospect.

The test lasted three full months. During this period, he produced fifteen to twenty quotes per month, a volume representative of his usual activity. He tested several approaches: a general conversational assistant for drafting descriptions, a language model integrated into his word processor, and prompts refined over the weeks to match his usual tone.

The evaluation criteria were concrete: actual writing time, chronometered for each quote. Number of corrections needed before sending. Client reactions, when measurable. And above all, a simple question asked after each quote sent: does this text sound like him?

AI for writing a quote was therefore not tested as a gadget, but as a daily work tool, with its real constraints: short deadlines, demanding clients, and a professional reputation to preserve with each submission.

What AI truly changed in the workflow

Over the three months, three effects became permanent. None were spectacular overnight, but all ended up modifying the way of working.

The end of the blank page

The first change isn't technical, it's psychological. Faced with a new prospect, this writer often wasted time even before starting to write. AI removed this initial block: a first version appears in a few seconds, even if imperfect. It serves as a starting point, not a final text. This simple fact reduced the initiation time for each quote, the time spent setting a structure before writing the first sentence.

Raw generation speed

Before AI, drafting a complete quote usually took between twenty and forty minutes, depending on the complexity of the task. With AI, the first version of a service description appears in one to two minutes. Over a month of fifteen quotes, this represents several hours freed up in the generation phase alone. This gain is real, measurable, and it was confirmed throughout the test period, including for unusual tasks.

Consistency of structures and formulations

Another notable effect: all generated quotes adopt the same architecture. Same order of sections, same level of detail, same phrasing for general terms and conditions. No more sloppy quotes on a Friday evening, with awkward wording or formatting oversights. This uniformity gave a more professional image to all documents sent, without extra effort for formal proofreading.

How much real time is saved each week?

This is the question that most interests a freelancer: the time saved, once everything is accounted for, including corrections.

The table reveals the real challenge. Raw drafting collapses, but proofreading increases, because a generated text requires more careful verification than a self-written text. For five to fifteen quotes per month, the real equation gives a net gain of about ten to fifteen minutes per quote, once corrections are accounted for. This is far from the promises of instant time savings often associated with AI. The gain exists, but it is two to three times more modest than the initial impression suggests.

What AI hasn't solved (and still costs you time)

Over three months, some problems never resolved themselves. They reappeared, almost identical, with each new series of quotes.

Pricing errors and forgotten services

This is the most costly pitfall. AI sometimes invents a price that doesn't exist in the writer's grids, or reuses an old amount mentioned in a previous exchange, without distinguishing what is still valid. It also forgets ancillary services, such as SEO proofreading or a revision included in the offer. In practice, this type of oversight occurred several times during the test, often for slightly atypical tasks. Each error not detected before sending can directly erode profit margins, or create a misunderstanding that is difficult to correct afterwards.

Inability to deeply understand each client

An effective quote takes into account the client's real budget, their previous requests, what they don't always dare to say. AI ignores all of this by default. It produces a generic text, adapted to a typical task, not to this specific client. Result: a technically correct quote, but one that lacks the nuance that would have reassured a hesitant client or convinced a rushed client.

A generic tone that isn't yours

For a writer, tone is a signature. It is often what distinguishes a standard offer from a quote that makes one want to sign. AI smooths this tone. It produces correct sentences, but interchangeable from one writer to another. Over time, some regular clients noticed this change, without always articulating it clearly: the quote seemed less personal, almost copied from a generic template.

Legal risks and compliance

Last point, often overlooked. AI regularly forgets mandatory mentions: offer validity period, payment terms, or clauses specific to the activity. It also does not integrate, without being explicitly asked, information on late payment penalties, which must at minimum correspond to three times the legal interest rate, accompanied by a flat-rate recovery indemnity of €40 per invoice. The payment deadline between professionals remains by default 30 days, unless otherwise agreed, with a legal ceiling of 60 days from the invoice date or 45 days end of month. These clauses are not decorative: responsibility for their presence remains entirely with the writer, never with the tool used to write the text.

Should every quote be fully proofread before sending?

The honest answer, after three months, is yes, almost systematically. Quotes that can be sent without full proofreading are rare: typically, an identical renewal for an existing client, with a task strictly similar to the previous one.

In all other cases, a complete review is necessary. Here are the points that deserve a quick check, with each submission:

• Do the prices exactly match the current grid, with no invented or obsolete amounts?
• Do all services discussed with the client appear in the description?
• Does the tone of the text resemble the one usually used with this client?
• Are the offer's validity period and payment terms present?
• Does the quote mention a reference or detail specific to the initial exchange with this prospect?

This checklist takes two to three minutes once you're used to it. It's short, but essential. Ignoring it, even once, is sometimes enough to send a quote with an incorrect price or a missing service.

How not to lose your client relationship in the process?

A quote is not just a commercial document. It is often the first written contact with a client, setting the tone for the entire collaboration. Used indiscriminately, a generated text can break this connection, especially with clients accustomed to a well-identified style.

Some clients perceive the impersonal without precisely naming it. They just feel that something has changed. To preserve authenticity without sacrificing time savings, several adjustments proved useful during the test. Always keep an opening sentence handwritten, never generated. Inject an element specific to the exchange with the client, a reference to a conversation or a detail from the brief. Read the text aloud before sending, to identify formulations that don't match the usual tone.

AI can build the skeleton of the quote. Personalisation, however, must remain the responsibility of the writer, for the passages that truly matter for the client relationship.

The true profitability calculation for a freelancer

The profitability calculation goes beyond simple writing time. It must include the initial investment: learning the tool, testing prompts, initial failures that required starting over. In the first weeks of the test, this learning time partly cancelled out the expected gain.

Then, the recurring cost, often underestimated, must be added: proofreading and verification, which are longer than with a self-written text. This cost never completely disappears, even after three months of practice.

Real savings depend primarily on volume. Below five quotes per month, the time invested in learning and verification often exceeds the gain obtained. Beyond ten to fifteen quotes per month, profitability becomes clear, because the time saved on raw drafting is repeated on a large scale.

Winning profiles are those who send many similar quotes, with limited variations from one client to another. Losing profiles are those whose every task is unique, where personalisation, in any case, takes precedence over generation speed.

For the invoicing part that follows the quote's signature, some tools like Quickote offer quote and invoice creation with a promised speed of under 60 seconds. Quickote génère vos factures au format Factur-X. Le raccordement à une Plateforme Agréée, nécessaire à l'émission obligatoire de septembre 2027, est en cours de finalisation.

Verdict: Has AI transformed your freelance life?

After three months, the assessment is nuanced, neither a miracle nor a disappointment. AI for writing a quote has genuinely changed the first step of the process: it eliminated the blank page and accelerated raw production. On this specific point, the change is clear and lasting.

What no AI has replaced, however, is the detailed knowledge of each client, the writer's own tone, and vigilance over prices and mandatory clauses. These elements remain, and will remain, human work.

The conditions for success are clear: a sufficient volume of quotes, a systematic proofreading checklist, and discipline to maintain a degree of manual personalisation on each submission. Before starting, the real question to ask is not "Will AI write my quotes?", but "Am I ready to proofread every text as if I had written it myself?"

In summary

AI for writing a quote speeds up raw drafting but shifts the time saved towards proofreading and verification. A net gain exists, especially beyond ten quotes per month, provided a systematic control checklist is maintained. Client personalisation and mandatory mentions remain entirely the responsibility of the writer.