[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news:jp-tech-smarthr-ai-native-dev-2026-06-05":3},{"meta":4,"markdown":63},{"type":5,"articleId":6,"slug":7,"title":8,"titleEn":9,"category":10,"summary":11,"publishedAt":12,"image":13,"tags":14,"vocabulary":17,"source":61},"news","news-jp-tech-smarthr-ai-native-dev-2026-06-05","jp-tech-smarthr-ai-native-dev-2026-06-05","SmartHR、AI前提で開発プロセスを作り直す実験を開始","SmartHR Starts an Experiment to Rebuild Development Around AI","tech-jp-culture","SmartHR said on June 4 that its attendance-management team began a three-month project in April to redesign development around AI rather than bolt AI onto an old workflow. The company says it is treating the effort as a hypothesis-driven experiment, with explicit quality rules and business approval to prioritize process learning over short-term feature throughput. The post offers a concrete look at how a Japanese software team is trying to operationalize AI-assisted development instead of using AI only as a coding helper.\n","2026-06-05T00:00:00Z","https:\u002F\u002Fimages.yamiyomi.com\u002Fnews\u002Fjp-tech-smarthr-ai-native-dev-2026-06-05.png",[15,16],"japan","ai",[18,23,28,33,37,41,45,49,53,57],{"word":19,"reading":20,"meaning":21,"level":22},"勤怠管理","きんたいかんり","attendance management","N1",{"word":24,"reading":25,"meaning":26,"level":27},"前提","ぜんてい","premise","N2",{"word":29,"reading":30,"meaning":31,"level":32},"開発","かいはつ","development","N3",{"word":34,"reading":35,"meaning":36,"level":22},"再設計","さいせっけい","redesign",{"word":38,"reading":39,"meaning":40,"level":32},"実験","じっけん","experiment",{"word":42,"reading":43,"meaning":44,"level":22},"仮説","かせつ","hypothesis",{"word":46,"reading":47,"meaning":48,"level":32},"評価","ひょうか","evaluation",{"word":50,"reading":51,"meaning":52,"level":27},"品質","ひんしつ","quality",{"word":54,"reading":55,"meaning":56,"level":22},"自律","じりつ","autonomy",{"word":58,"reading":59,"meaning":60,"level":27},"展開","てんかい","rollout",{"url":62},"https:\u002F\u002Ftech.smarthr.jp\u002Fentry\u002F2026\u002F06\u002F04\u002F094730","\n::heading\nSmartHR、AI[前提]{ぜんてい:premise:N1}で[開発]{かいはつ:development:N4}プロセスを[作り直す]{つくりなおす:rebuild:N3} [実験]{じっけん:experiment:N3}を[開始]{かいし:start:N4}\n\n#en\nSmartHR Starts an Experiment to Rebuild Development Around AI\n::\n\n::para\nSmartHRは6月4日、[勤怠管理]{きんたいかんり:attendance management:N1}チームがAIを[前提]{ぜんてい:premise:N1}にした[開発]{かいはつ:development:N4}プロセスへの[移行]{いこう:shift:N2}を[進めている]{すすめている:is advancing:N3}と[紹介]{しょうかい:introduced:N2}した。[記事]{きじ:article:N3}によると、2026年4月から3か[月]{げつ:months:N5}の[計画]{けいかく:plan:N4}として[始まった]{はじまった:started:N4}という。\n\n#en\nOn June 4, SmartHR introduced an effort by its attendance-management team to shift development toward an AI-first process. According to the post, the plan began in April 2026 as a three-month project.\n::\n\n::para\n[狙い]{ねらい:aim:N2}は、AIツールを[既存]{きそん:existing:N1} [工程]{こうてい:workflow:N3}に[足す]{たす:add:N4}ことではなく、AIがいることを[前提]{ぜんてい:premise:N1}に[仕事]{しごと:work:N4}の[流れ]{ながれ:flow:N3}そのものを[再設計]{さいせっけい:redesign:N2}することだとされる。SmartHRはこの[考え方]{かんがえかた:approach:N4}を「AIネイティブ[開発化]{かいはつか:shift to AI-native development:N3}」と[呼んでいる]{よんでいる:calls:N3}。\n\n#en\nThe stated goal is not to add AI tools to an old workflow, but to redesign the flow of work itself on the assumption that AI is always present. SmartHR calls this approach a shift to AI-native development.\n::\n\n::heading\n[変える]{かえる:change:N3} [理由]{りゆう:reason:N3}\n\n#en\nWhy the Team Wants to Change\n::\n\n::para\n[記事]{きじ:article:N3}では、チームは[以前]{いぜん:before:N4}からAI[活用]{かつよう:use:N3}に[積極的]{せっきょくてき:active:N2}だったものの、[期待]{きたい:expectation:N3}したほどの[速度]{そくど:speed:N3}は[出なかった]{でなかった:did not appear:N5}と[説明]{せつめい:explained:N3}している。[従来]{じゅうらい:conventional:N1}の[役割分担]{やくわりぶんたん:division of roles:N2}や[確認手順]{かくにんてじゅん:review steps:N2}が、そのままでは[足かせ]{あしかせ:constraint:N4}になったとされる。\n\n#en\nThe post says the team had already been active in using AI, but the expected speed gains did not fully appear. It explains that conventional role splits and review steps became constraints when left unchanged.\n::\n\n::para\nそのため、SmartHRは1つの[チーム]{ちーむ:team}を[パイロット]{ぱいろっと:pilot}にして、[機能開発]{きのうかいはつ:feature development:N3}よりも[実験]{じっけん:experiment:N3}を[優先]{ゆうせん:prioritize:N3}してよいと[決めた]{きめた:decided:N3}。[四半期]{しはんき:quarter:N3}の[終わり]{おわり:end:N4}までに、[学び]{まなび:learning:N5}をチーム[全体]{ぜんたい:whole team:N3}へ[広げられる]{ひろげられる:can spread:N4}かどうかで[評価]{ひょうか:evaluate:N1}するとしている。\n\n#en\nSmartHR therefore chose one pilot team and decided it could prioritize experimentation over feature delivery. It says the project will be judged by whether the lessons can spread across the broader team by the end of the quarter.\n::\n\n::heading\n[実験]{じっけん:Experiment:N3}の[進め方]{すすめかた:how it is run:N3}\n\n#en\nHow the Experiment Is Being Run\n::\n\n::para\nSmartHRは、[背景]{はいけい:background:N3}、[実施内容]{じっしないよう:what will be done:N1}、[確かめたい]{たしかめたい:wants to verify:N3} [仮説]{かせつ:hypothesis:N1}、リスク[仮説]{かせつ:hypothesis:N1}、そして[評価観点]{ひょうかかんてん:evaluation criteria:N1}をそろえたテンプレートで、[実験]{じっけん:experiment:N3}を[回す]{まわす:run:N3} [方針]{ほうしん:policy:N2}だという。[勘]{かん:gut feeling:N1}ではなく、[比較可能]{ひかくかのう:comparable:N1}な[形]{かたち:form:N3}で[学習]{がくしゅう:learning:N4}を[蓄積]{ちくせき:accumulate:N1}したいようだ。\n\n#en\nThe company says it is using a template that captures background, actions, target hypotheses, risk hypotheses, and evaluation criteria. The point appears to be to accumulate learning in a comparable form rather than rely on gut feeling.\n::\n\n::para\n[記事]{きじ:article:N3}は、[実験]{じっけん:experiment:N3}と[評価]{ひょうか:evaluation:N1}の[循環]{じゅんかん:cycle:N1}を[短く]{みじかく:short:N2} [回す]{まわす:run:N3}ことが[重要]{じゅうよう:important:N3}だと[示している]{しめしている:indicates:N3}。AIを[導入]{どうにゅう:introduce:N2}しただけではなく、[使い方]{つかいかた:way of using:N4}と[判断基準]{はんだんきじゅん:decision criteria:N1}まで[揃える]{そろえる:align:N1}ことが[肝心]{かんじん:crucial:N1}だと[見られる]{みられる:is seen:N5}。\n\n#en\nThe article suggests it is important to keep the experiment-and-evaluation loop short. The implication is that introducing AI is not enough unless usage patterns and decision criteria are aligned as well.\n::\n\n::para\n[品質]{ひんしつ:quality:N4}については、[内部品質]{ないぶひんしつ:internal quality:N3}も[外部品質]{がいぶひんしつ:external quality:N3}も、[人間]{にんげん:humans:N5}が[担保]{たんぽ:guarantee:N1}してきた[水準]{すいじゅん:standard:N2}を[落とさない]{おとさない:not lower:N3}ことを[前提]{ぜんてい:premise:N1}にしているという。[速度]{そくど:speed:N3}のために[品質]{ひんしつ:quality:N4}を[犠牲]{ぎせい:sacrifice:N1}にしない[線引き]{せんびき:boundary line:N2}を、[最初]{さいしょ:the start:N3}から[明文化]{めいぶんか:make explicit:N3}した[形]{かたち:form:N3}だ。\n\n#en\nOn quality, the post says both internal and external quality should stay at the level previously guaranteed by humans. In effect, SmartHR is drawing an explicit line from the start against trading quality for speed.\n::\n\n::heading\n[日本]{にほん:Japan:N5}の[開発組織]{かいはつそしき:engineering organization:N1}への[示唆]{しさ:suggestion:N1}\n\n#en\nWhat This Means for Japanese Dev Teams\n::\n\n::callout\nAIが[書く]{かく:write:N5}コードの[量]{りょう:volume:N2}が[増えても]{ふえても:even if it increases:N3}、[問題設定]{もんだいせってい:problem framing:N2}、[評価]{ひょうか:evaluation:N1}、[統合]{とうごう:integration:N1}、そして[責任]{せきにん:responsibility:N3}の[置き方]{おきかた:how responsibility is assigned:N3}は[人間側]{にんげんがわ:human side:N3}に[残る]{のこる:remain:N3}。そのため、AIネイティブ[開発]{かいはつ:development:N4}は[単なる]{たんなる:mere:N3}ツール[導入]{どうにゅう:introduction:N2}よりも、[組織設計]{そしきせっけい:organizational design:N1}の[話]{はなし:matter:N5}に[近い]{ちかい:closer:N4}といえる。\n\n#en\nEven if the volume of AI-written code grows, problem framing, evaluation, integration, and responsibility still remain on the human side. That makes AI-native development closer to organizational design than to simple tool adoption.\n::\n\n::para\nSmartHRの[事例]{じれい:case:N3}が[興味深い]{きょうみぶかい:interesting:N1}のは、[生産性]{せいさんせい:productivity:N3}だけでなく、[自律的]{じりつてき:autonomous:N2}に[判断]{はんだん:decide:N3}できる[範囲]{はんい:scope:N1}をどう[広げる]{ひろげる:expand:N4}かまで[含めて]{ふくめて:including:N2}いる[点]{てん:point:N3}だ。AIの[得意]{とくい:strength:N3}な[部分]{ぶぶん:parts:N3}と[人間]{にんげん:humans:N5}が[持つ]{もつ:hold:N4} [文脈理解]{ぶんみゃくりかい:context understanding:N1}を、どう[組み合わせる]{くみあわせる:combine:N3}かが[問われている]{とわれている:is being tested:N4}。\n\n#en\nWhat makes the SmartHR case notable is that it is not only about productivity, but also about how far teams can expand autonomous decision-making. The deeper question is how to combine the parts AI is good at with the contextual understanding humans still provide.\n::\n\n::para\n[日本]{にほん:Japan:N5}のソフトウェア[企業]{きぎょう:software companies:N1}でも、AI[活用]{かつよう:use:N3}は[個人]{こじん:individual:N2}の[工夫]{くふう:ingenuity:N3}から[組織的]{そしきてき:organizational:N1}な[運用]{うんよう:operations:N4}へ[移りつつある]{うつりつつある:is shifting:N2}と[見られる]{みられる:is seen:N5}。SmartHRがこの[実験]{じっけん:experiment:N3}の[結果]{けっか:results:N1}を[続報]{ぞくほう:follow-up:N3}でどう[示す]{しめす:show:N3}のかは、ほかの[開発現場]{かいはつげんば:development sites:N3}にとっても[参考]{さんこう:reference:N3}になりそうだ。\n\n#en\nAcross Japanese software companies, AI use appears to be shifting from individual improvisation toward organizational operation. How SmartHR reports the results of this experiment in follow-up posts will likely become useful reference material for other teams.\n::\n"]