LinkedIn Content Strategy for AI Consultants for SMEs | Netwarm
LinkedIn playbook · Case study
LinkedIn content strategy for AI consultants: how a Paris outsourced AI lead reached 92 SME decision-makers in 3 weeks
We read every reaction on 30 posts from a Paris outsourced AI lead for SMEs. 648 people, 92 founders, CEOs and directors. Here is what brings buyers, what brings peers, and the 5 posts we’d publish next.
Max Sawyer · Netwarm··11 min read
The creator in this case studyA Paris outsourced AI lead for SMEs
Followers14,447
Posts~10 a week
SellsOutsourced AI leadership for SMEs: AI roadmap, agents, internal tools and automations, plus coaching for managers and CEOs
PeriodOct 12 – Nov 1
Ideal client: founders, CEOs and IT directors of French SMEs who have started with AI but don’t know where it is taking them.The challenge: The consultant wrote it: “Scaler sans recruter, intégrateur Claude, expert IA, automatisation, systèmes… J’ai tout essayé comme tagline.” Ten posts a week bring a big AI crowd: freelancers, automation experts, agencies. Do they also reach the SME owners who hire an outsourced AI lead?
Key takeaways
These numbers are from 1 November. Your audience has already moved.
New SME decision-makers turn Hot every week and others go quiet. Run your free scan to see today’s version, with the names.
1 reader in 7 is an SME decision-maker.92 founders, CEOs, owners and directors reacted in 3 weeks, 14.2% of the 648 people who reacted. AI freelancers, consultants, agencies and trainers are 47.8%.
The free guides beat the personal posts.The guides brought 15.5% decision-makers, the personal and tool posts 9.8%: 1.6x more per reaction.
4 decision-makers keep coming back.They reacted to 3 posts or more in 3 weeks. One of them reacted to 5 of the 30.
01
Who actually reads you, and what the LinkedIn algorithm changed
The consultant posts about 10 times a week, in French: AI news decoded, free guides of 35 to 68 slides, the method used with SMEs, and personal notes on tools and podcasts.
The LinkedIn algorithm shows a post to the people who reacted to the last ones, then to people like them. When AI freelancers and automation experts react the most, the feed keeps reaching more of them. Reach can’t tell whether the SME owners who hire an outsourced AI lead are in it. Who reacted can. We looked at every person who reacted between 12 October and 1 November, and sorted them by role.
30posts in 3 weeks
648people reacted
92SME decision-makers
4came back 3+ times
Audience by roleThe 648 people who reacted, by roleFree
AI freelancers, consultants, agencies, trainers & assistants310 · 47.8%
Founders, CEOs, owners & directors92 · 14.2%
Everyone else246 · 38%
Peers lead: 310 AI and automation freelancers, consultants, agencies, trainers and assistants. Almost half of everyone who reacts. They comment and share, and some send work, but they sell what the consultant sells.
The buyers are there too: 92 founders, CEOs, owners, CTOs and directors, from tech and digital firms to medical device makers, insurance brokers and building companies. 1 reader in 7.
SeniorityAll people who reactedFree
Founder, owner, C-level12.8%
Director, VP3.1%
Manager9.1%
Independent9.6%
Other65.4%
IndustriesSME decision-makers onlyNetwarm
Tech, digital and data41.3%
Hidden in preview00%
Hidden in preview00%
Hidden in preview00%
See all industries and company sizes in your free scan.
Core audience4 SME decision-makers engaged 3+ times · 1 engaged 5+ timesNetwarm
4 SME decision-makers keep coming back. One of them reacted to 5 of the 30 posts.See your regulars
Engager company sizeEveryone who reacted, by company sizeNetwarm
Hidden in preview00%
Hidden in preview00%
Hidden in preview00%
Hidden in preview00%
Hidden in preview00%
See which company sizes your SME decision-makers work at.See company sizes
02
What to post on LinkedIn to reach SME decision-makers
Most AI creators judge a post by its reactions. A consultant who sells to SME owners should judge it by who reacted.
We tagged each of the 30 posts by topic, then measured the share of founders, CEOs and directors among the people who reacted, and how many of them each post reached.
Your topicsWhat your 30 posts were aboutFree
30posts
AI news decoded12
AI in SMEs, the method7
Personal and tools6
Free guides5
Topics that attract SME decision-makersShare of each topic’s reactions from SME decision-makersNetwarm
Free guides15.5%
AI news decoded12.7%
AI in SMEs, the method10.5%
Personal and tools9.8%
One topic brings 1.6x more SME decision-makers than another. See which.See which topics
The free guides bring 1.6x more SME decision-makers per reaction than the personal and tool posts: 15.5% against 9.8%. AI news reaches the most per post: 4.8.
Posts that brought the most SME decision-makersShare of SME decision-makers in each post’s reactions1 of 5 open
#1Ben Affleck sold his AI startup to Netflix22.9%
#2I never give my AI agents a password18.2%
#3The guide to build your company’s internal OS with Claude Code17.3%
#4A designer shares his AI skills for free15%
#5Skills are the processes of AI10.8%
For comparison, the podcast post brought 8.3% decision-makers, and the posts that re-shared a guide ten days later got 8 reactions each. The guide works when it is new and asked for, not when it is posted as a link.
6 rules for LinkedIn engagement from SME decision-makers
These rules come from what worked on the consultant’s own profile in October, and from the best-performing posts by French AI consultants and SME advisors on LinkedIn since August.
1
Give the guide for a comment. A guide built from real projects is the proof an SME owner needs before a call. Asking for a comment turns readers into a list of people to follow up with.On this profile: the internal OS guide got 173 reactions and 17.3% decision-makers. The 68-slide SME guide of 29 September got 288 likes and 509 comments. The same guides re-shared as a link got 8 reactions each.
Tell the news through what it changes for an SME. A big AI story gets read. The line that turns it into a lesson for a business owner is what makes owners react.On this profile: the Ben Affleck post, about business expertise beating the tech, brought 22.9% decision-makers. AI news reached 4.8 decision-makers per post, the most of any topic.
Tell what you refused to build. Saying no to a client request shows judgment. That is what an owner buys from an outsourced AI lead, more than the build itself.Across the niche: an AI ROI founder’s post on refusing to automate a client’s bad-news emails got 114 likes from an 8.2K-follower account.
Say where the agent got it wrong. Owners fear the confident mistake. A real case where the agent missed something, and what you changed, builds more trust than a demo.Across the niche: an accountant’s post on a Claude agent that missed a client’s 38,000 € bad debt got 106 likes and 40 comments, from a 1.7K-follower account.
Show the real build, hour by hour. An owner can’t judge an agent from a concept. The timeline, the steps and what you checked by hand make it real.Across the niche: a tech MD’s post on building a talk-to-video machine with Claude Code, with every timestamp, got 125 likes. On this profile, the software factory post explained the idea without showing it: 9.1% decision-makers.
Keep the tool talk for the peers. Podcasts, browsers and model prices interest other AI freelancers. Owners want to know what changes in their company.On this profile: the personal and tool posts brought 9.8% decision-makers. The model price post (GPT-6.1 Terra) and the Lovable post brought none.
When SME decision-makers read: the best time to post on LinkedIn for AI consultants
We compared the share of SME decision-makers by the day and the hour (Paris time) each post was published. Only hours with 3 or more posts are shown.
Share of SME decision-makers by posting dayAmong the people who reacted to each postFree
15.1%
Mon
13.7%
Tue
9.7%
Wed
6.8%
Thu
11.6%
Fri
8.3%
Sat
15.6%
Sun
Best time to post, by hourShare of SME decision-makers among people who reactedFree
12.5%
8:00
9.1%
10:00
16.9%
14:00
Tuesday brought the most decision-makers per post (8.4), with the internal OS guide and the designer skills post. Sunday (15.6%) and Monday (15.1%) brought the highest shares. Thursday was the weakest day (6.8%).
The 8:00 posts brought 12.5% decision-makers and 4.9 per post, the 10:00 posts 9.1% and 1.4 per post: the second post of the day reaches far fewer people. Post once a day at 8:00, and keep Tuesday for the guides.
05
5 posts ready to publish to target SME decision-makers
Each post follows the rules above, in the consultant’s voice and in French, like the rest of the feed. Swap in your own details before posting: the real projects and the real numbers are what make them land.
1Ready to postMethod + lead magnet
Où en est ta boîte sur les 4 cycles de l’IA ?Rule 1 and the lead magnet: the 4-cycle method from the 1 November post, turned into a one-page diagnostic and offered for a comment, on a Tuesday at 8:00 like the guides.
Hook
3 DSI et 8 dirigeants de PME. Tous avaient lancé des initiatives IA. Aucun ne savait où elles les menaient.C’est ce que j’ai entendu le mois dernier. Et c’est presque toujours le même schéma : des licences, une formation, quelques usages qui plaisent. Puis plus rien ne bouge.La méthode que j’applique tient en 4 cycles, qui tournent en même temps :1. La vision : un cap qui tient sur une page, et une feuille de route qu’on révise à chaque tour
2. La productivité perso : chacun gagne du temps sur ce qu’il fait déjà, les managers en premier
3. La productivité d’entreprise : chaque process passe par étapes de l’existant au réinventé
4. La R&D : le temps libéré sert à tester une offre, un service, un marchéLa plupart des PME que je rencontre sont bien avancées sur le 2. Presque aucune n’a commencé le 1.J’en ai fait une grille d’une page : 12 questions, un score par cycle, et le premier chantier à lancer selon ton résultat.Et une version pour ton IA : tu lui décris ta boîte, elle remplit la grille avec toi.Commente CAP et je te l’envoie.Back to work ! ✌️
Visual · DocumentLa grille elle-même : les 4 cycles en colonnes, 3 questions chacun, une case score en bas, aux couleurs de tes guides.Check before postingÉcris la grille (12 questions) avant de publier, vérifie que le constat « avancées sur le 2, rien sur le 1 » correspond à tes échanges, et réponds à chaque commentaire le jour même.
Le projet IA que j’ai refuséRule 3: tell what you refused to build. Shaped like Chloé Roumengas’s refusal post, on a Sunday at 8:00, the best share of the week.
Hook
Un dirigeant m’a demandé de reconstruire son CRM avec Claude Code. J’ai dit non.Sa PME : 30 personnes, un CRM du marché, et une équipe commerciale qui ne le remplit pas.Techniquement, c’était faisable. Quelques jours de travail.Mais sur la durée de vie d’un logiciel, 60 à 90 % du coût part en maintenance. Et le problème n’était pas l’outil.Quand on a regardé de plus près :
- les commerciaux ressaisissaient chaque devis à la main
- les relances vivaient dans un fichier Excel à part
- personne ne savait quelle colonne faisait foiUn nouveau CRM aurait eu exactement le même problème. En plus cher à maintenir.Alors on a fait autre chose :
- on a branché un agent sur le CRM existant, qui le remplit après chaque rendez-vous
- on a supprimé le fichier des relances
- chaque fiche déclenche maintenant une actionLe CRM n’a pas changé. L’équipe s’en sert.Avec l’IA, la question n’est plus « est-ce qu’on peut le construire ? ». C’est presque toujours oui.C’est « est-ce que ça change le travail de l’équipe ? ».Bon dimanche ! ✌️
Visual · Single imageDeux colonnes sur fond blanc : « Ce qu’on m’a demandé » (un nouveau CRM) et « Ce qu’on a fait » (3 lignes), en noir et une couleur d’accent.Check before postingUtilise un vrai projet et ses vrais chiffres (taille de l’équipe, outils). Ne garde rien qui permette de reconnaître le client.
Amazon bloque les agents. Ton site, lui, est-il prêt ?Rule 2: a news story told through what it changes for an SME. The facts come from the 21 October post, with the owner’s angle that made the Ben Affleck post work.
Hook
Depuis le 14 octobre, Amazon bloque les agents d’achat. Shopify, lui, leur ouvre ses boutiques.Deux géants, deux paris opposés :- Delta bloque les agents qui veulent acheter des billets
- Yelp n’accepte que ceux qui paient sa licence de données
- JPMorgan et Schwab bloquent leur accès aux comptes clientsEn face :- Meta signe avec Walmart, Gap, Sephora et Best Buy pour que son agent puisse y acheter
- Shopify ouvre ses boutiques aux agentsPour une PME, la question arrive plus vite qu’on ne le croit.Si tu vends en ligne : un agent sait-il trouver tes prix, tes stocks, tes conditions de livraison ? Ou est-ce que tout est caché dans des PDF et des pages qui changent chaque semaine ?Si tu vends en B2B : un acheteur qui demande à son IA de comparer trois fournisseurs, est-ce qu’elle trouve les informations sur toi ?On passe des années à rendre un site lisible pour Google. Le prochain lecteur, c’est un agent.Tu laisserais un agent acheter chez toi ?Back to work ! ✌️
Visual · Single imageDeux colonnes de logos : « Ils bloquent les agents » (Amazon, Delta, Yelp) et « Ils les accueillent » (Meta, Shopify), avec la question en bas.Check before postingRevérifie chaque fait (dates, entreprises) le jour de la publication : l’actualité des agents bouge vite.
L’agent avait raison sur tout. Sauf sur l’essentiel.Rule 4: say where the agent got it wrong, and what changed. Shaped like Odélia Amar’s pre-audit agent post (106 likes, 40 comments), on a Tuesday at 8:00.
Hook
L’agent avait raison sur tout. Sauf sur l’essentiel.Chez un client, on a construit un agent qui prépare les relances de devis. Il lit le CRM, retrouve le devis, rédige un mail dans le ton de la boîte.Les premières semaines : parfait. Les commerciaux validaient presque tout sans retoucher.Puis un jour, il prépare une relance impeccable pour un client… qui avait annulé la veille. Par téléphone.Rien dans le CRM. Rien dans les mails. L’agent a fait exactement ce qu’on lui avait demandé, avec ce qu’on lui avait donné.Le commercial l’a vue avant l’envoi. Parce que c’est lui qui valide.Ce qu’on a changé :
1. Chaque appel important laisse une ligne dans le CRM, dictée en 20 secondes
2. L’agent signale quand il n’a aucune trace d’échange depuis 30 jours
3. Il ne relance jamais seul, il prépareUne IA ne se trompe pas dans ses calculs. Elle se trompe sur ce qu’on ne lui a pas dit.Ton contexte client, il est où ? Dans tes outils, ou dans la tête de tes commerciaux ?Back to work ! ✌️
Visual · Single imageUne capture floutée d’un brouillon de mail de relance, avec en rouge « Client annulé hier par téléphone ».Check before postingUtilise un vrai cas où un agent s’est trompé, avec ce que tu as réellement changé. Si tu n’en as pas, garde la structure et raconte un cas que tu as vu chez un confrère, en le disant.
Une journée dans ma software factoryRule 5: show the real build, hour by hour. Takes the software factory idea of the 31 October post and tells it like Didier Girard’s timestamped build.
Hook
9 h 10, un client m’envoie un vocal : « Tu peux ajouter un export des devis ? » 16 h 40, c’est en production.Entre les deux, ma software factory. Une chaîne d’agents IA qui produit du logiciel de bout en bout. Un humain valide aux étapes clés. L’humain, c’est moi.Le déroulé :9 h 20 → un agent transforme le vocal en spécification. Je la relis, je corrige deux points.
10 h 05 → un autre écrit le code et les tests.
11 h 30 → je teste sur une copie de la base du client. Un format de date ne va pas, je le signale.
14 h 00 → l’agent corrige, relance les tests, met à jour la documentation.
16 h 15 → je valide la mise en production.
16 h 40 → le client reçoit le message : c’est en ligne.Mon temps sur la journée : moins d’une heure.Ce qui ne se délègue pas : comprendre ce que le client veut vraiment, vérifier sur ses vraies données, et décider de ce qui part en production.Demain, toutes les entreprises auront une partie technique à maintenir. La question n’est pas de savoir coder. C’est d’avoir un cadre pour que ça tienne.Back to work ! ✌️
Visual · Single imageUne frise horaire de 9 h à 17 h avec les 6 étapes, les moments « humain » en couleur et les moments « agents » en gris.Check before postingRemplace les heures et la fonctionnalité par une vraie journée tirée de tes logs. Garde « moins d’une heure » seulement si c’est vrai.
Free forever: unlimited scheduling through LinkedIn’s official API, with images, video and PDF carousels.
06
LinkedIn lead generation: lead magnets that work to target SME decision-makers
Reactions show interest. A download shows intent. The quickest way to find out which SME owners are serious is to give them something useful and see who asks for it.
Among French AI consultants and SME advisors, these formats get the most requests:
Format 1The guide in slidesA long guide built from real projects, with a version to give to your own AI. Sent to everyone who comments.
Format 2The prompt to copyOne ready-made prompt that does a real job for the reader, sent for a comment.
Format 3The live session and its replayShow a real case live for an hour. Those who can’t come comment to get the replay.
Our pick for you: “The 4-cycle AI diagnostic”Built from the consultant’s 4-cycle method (vision, personal productivity, company productivity, R&D): 12 questions, a score per cycle and the first project to launch, on one page, with a version for the reader’s own AI. Offer it in post 1, on a Tuesday at 8:00.With a Reveal link, every person who downloads it is shown by name, with their role and how warm they are. Your first Reveal link is free.Create your free Reveal link
07
From reactions to clients
Content brings SME owners to the profile. It doesn’t say when to reach out. Most never comment, so they stay invisible, and a CEO reading every AI post in October may have hired someone else by January.
On LinkedIn, the right moment to start a conversation is when someone keeps coming back. We call that Hot: 5 or more points over 90 days, where a reaction counts 1, a repost 2, a download or a reply 4.
Hot right now1 SME decision-makers · not contacted yetNetwarm
1 SME decision-makers are Hot. See who to message first.See who’s Hot
First messageAn opener for each Hot person, based on the posts they readNetwarm
Bonjour Marc, vous lisez mes posts sur l’IA en PME depuis quelques semaines. C’est un sujet sur votre table en ce moment ?Bonjour Claire, j’ai vu que le guide sur l’OS interne vous avait parlé. Vous avez un fichier Excel qui fait mal à l’équipe ?
A first line for each of them, ready to send.See the openers
Netwarm sends nothing for you. It tells you who and when, and you write the message. It also works inside Claude and ChatGPT: ask “Who’s Hot today?” and get the list.
08
FAQ
Here, the free guides brought the highest share of SME decision-makers (15.5%), and AI news told through a business lesson brought up to 22.9%. Personal and tool posts brought the fewest (9.8%).
For this consultant, Tuesday brought the most decision-makers per post (8.4), Sunday (15.6%) and Monday (15.1%) the highest shares. Thursday was the weakest day (6.8%).
This consultant posts about 10 times a week. The 8:00 posts reached 4.9 decision-makers each, the 10:00 posts 1.4: one strong post a day beats two.
Look at who reacts, not how many. Almost half of the people who react here sell AI services themselves. Group reactions by role and follow the people who keep coming back. A free Netwarm scan does this for you.
Yes. Netwarm only reads public reactions to your own posts. It never logs in as you, never asks for your password, and never messages anyone for you.
How we made this playbook
We read the public reactions on 30 posts published between 12 October and 1 November 2026, up to the first 100 people per post (881 of the 1,261 reactions), and sorted the 648 people by role. Comments were not counted, so engagement figures are a floor.
Niche rules come from top posts by French AI consultants, AI trainers and SME advisors since August. Role groups are approximate; some founders also sell AI services. A Netwarm scan matches people to your own ideal client.
